1FFMPEG-FILTERS(1) FFMPEG-FILTERS(1)
2
6 ffmpeg-filters - FFmpeg filters
7
9 This document describes filters, sources, and sinks provided by the
10 libavfilter library.
11
13 Filtering in FFmpeg is enabled through the libavfilter library.
14 In libavfilter, a filter can have multiple inputs and multiple outputs.
16 To illustrate the sorts of things that are possible, we consider the
17 following filtergraph.
18 [main]
20 input --> split ---------------------> overlay --> output
21 | ^
22 |[tmp] [flip]|
23 +-----> crop --> vflip -------+
24 This filtergraph splits the input stream in two streams, then sends one
26 stream through the crop filter and the vflip filter, before merging it
27 back with the other stream by overlaying it on top. You can use the
28 following command to achieve this:
29 ffmpeg -i INPUT -vf "split [main][tmp]; [tmp] crop=iw:ih/2:0:0, vflip [flip]; [main][flip] overlay=0:H/2" OUTPUT
31 The result will be that the top half of the video is mirrored onto the
33 bottom half of the output video.
34 Filters in the same linear chain are separated by commas, and distinct
36 linear chains of filters are separated by semicolons. In our example,
37 crop,vflip are in one linear chain, split and overlay are separately in
38 another. The points where the linear chains join are labelled by names
39 enclosed in square brackets. In the example, the split filter generates
40 two outputs that are associated to the labels [main] and [tmp].
41 The stream sent to the second output of split, labelled as [tmp], is
43 processed through the crop filter, which crops away the lower half part
44 of the video, and then vertically flipped. The overlay filter takes in
45 input the first unchanged output of the split filter (which was
46 labelled as [main]), and overlay on its lower half the output generated
47 by the crop,vflip filterchain.
48 Some filters take in input a list of parameters: they are specified
50 after the filter name and an equal sign, and are separated from each
51 other by a colon.
52 There exist so-called source filters that do not have an audio/video
54 input, and sink filters that will not have audio/video output.
55
57 The graph2dot program included in the FFmpeg tools directory can be
58 used to parse a filtergraph description and issue a corresponding
59 textual representation in the dot language.
60 Invoke the command:
62 graph2dot -h
64 to see how to use graph2dot.
66 You can then pass the dot description to the dot program (from the
68 graphviz suite of programs) and obtain a graphical representation of
69 the filtergraph.
70 For example the sequence of commands:
72 echo <GRAPH_DESCRIPTION> | \
74 tools/graph2dot -o graph.tmp && \
75 dot -Tpng graph.tmp -o graph.png && \
76 display graph.png
77 can be used to create and display an image representing the graph
79 described by the GRAPH_DESCRIPTION string. Note that this string must
80 be a complete self-contained graph, with its inputs and outputs
81 explicitly defined. For example if your command line is of the form:
82 ffmpeg -i infile -vf scale=640:360 outfile
84 your GRAPH_DESCRIPTION string will need to be of the form:
86 nullsrc,scale=640:360,nullsink
88 you may also need to set the nullsrc parameters and add a format filter
90 in order to simulate a specific input file.
91
93 A filtergraph is a directed graph of connected filters. It can contain
94 cycles, and there can be multiple links between a pair of filters. Each
95 link has one input pad on one side connecting it to one filter from
96 which it takes its input, and one output pad on the other side
97 connecting it to one filter accepting its output.
98 Each filter in a filtergraph is an instance of a filter class
100 registered in the application, which defines the features and the
101 number of input and output pads of the filter.
102 A filter with no input pads is called a "source", and a filter with no
104 output pads is called a "sink".
105 Filtergraph syntax
107 A filtergraph has a textual representation, which is recognized by the
108 -filter/-vf/-af and -filter_complex options in ffmpeg and -vf/-af in
109 ffplay, and by the avfilter_graph_parse_ptr() function defined in
110 libavfilter/avfilter.h.
111 A filterchain consists of a sequence of connected filters, each one
113 connected to the previous one in the sequence. A filterchain is
114 represented by a list of ","-separated filter descriptions.
115 A filtergraph consists of a sequence of filterchains. A sequence of
117 filterchains is represented by a list of ";"-separated filterchain
118 descriptions.
119 A filter is represented by a string of the form:
121 [in_link_1]...[in_link_N]filter_name@id=arguments[out_link_1]...[out_link_M]
122 filter_name is the name of the filter class of which the described
124 filter is an instance of, and has to be the name of one of the filter
125 classes registered in the program optionally followed by "@id". The
126 name of the filter class is optionally followed by a string
127 "=arguments".
128 arguments is a string which contains the parameters used to initialize
130 the filter instance. It may have one of two forms:
131 • A ':'-separated list of key=value pairs.
133 • A ':'-separated list of value. In this case, the keys are assumed
135 to be the option names in the order they are declared. E.g. the
136 "fade" filter declares three options in this order -- type,
137 start_frame and nb_frames. Then the parameter list in:0:30 means
138 that the value in is assigned to the option type, 0 to start_frame
139 and 30 to nb_frames.
140 • A ':'-separated list of mixed direct value and long key=value
142 pairs. The direct value must precede the key=value pairs, and
143 follow the same constraints order of the previous point. The
144 following key=value pairs can be set in any preferred order.
145 If the option value itself is a list of items (e.g. the "format" filter
147 takes a list of pixel formats), the items in the list are usually
148 separated by |.
149 The list of arguments can be quoted using the character ' as initial
151 and ending mark, and the character \ for escaping the characters within
152 the quoted text; otherwise the argument string is considered terminated
153 when the next special character (belonging to the set []=;,) is
154 encountered.
155 A special syntax implemented in the ffmpeg CLI tool allows loading
157 option values from files. This is done be prepending a slash '/' to the
158 option name, then the supplied value is interpreted as a path from
159 which the actual value is loaded. E.g.
160 ffmpeg -i <INPUT> -vf drawtext=/text=/tmp/some_text <OUTPUT>
162 will load the text to be drawn from /tmp/some_text. API users wishing
164 to implement a similar feature should use the
165 "avfilter_graph_segment_*()" functions together with custom IO code.
166 The name and arguments of the filter are optionally preceded and
168 followed by a list of link labels. A link label allows one to name a
169 link and associate it to a filter output or input pad. The preceding
170 labels in_link_1 ... in_link_N, are associated to the filter input
171 pads, the following labels out_link_1 ... out_link_M, are associated to
172 the output pads.
173 When two link labels with the same name are found in the filtergraph, a
175 link between the corresponding input and output pad is created.
176 If an output pad is not labelled, it is linked by default to the first
178 unlabelled input pad of the next filter in the filterchain. For
179 example in the filterchain
180 nullsrc, split[L1], [L2]overlay, nullsink
182 the split filter instance has two output pads, and the overlay filter
184 instance two input pads. The first output pad of split is labelled
185 "L1", the first input pad of overlay is labelled "L2", and the second
186 output pad of split is linked to the second input pad of overlay, which
187 are both unlabelled.
188 In a filter description, if the input label of the first filter is not
190 specified, "in" is assumed; if the output label of the last filter is
191 not specified, "out" is assumed.
192 In a complete filterchain all the unlabelled filter input and output
194 pads must be connected. A filtergraph is considered valid if all the
195 filter input and output pads of all the filterchains are connected.
196 Libavfilter will automatically insert scale filters where format
198 conversion is required. It is possible to specify swscale flags for
199 those automatically inserted scalers by prepending "sws_flags=flags;"
200 to the filtergraph description.
201 Here is a BNF description of the filtergraph syntax:
203 <NAME> ::= sequence of alphanumeric characters and '_'
205 <FILTER_NAME> ::= <NAME>["@"<NAME>]
206 <LINKLABEL> ::= "[" <NAME> "]"
207 <LINKLABELS> ::= <LINKLABEL> [<LINKLABELS>]
208 <FILTER_ARGUMENTS> ::= sequence of chars (possibly quoted)
209 <FILTER> ::= [<LINKLABELS>] <FILTER_NAME> ["=" <FILTER_ARGUMENTS>] [<LINKLABELS>]
210 <FILTERCHAIN> ::= <FILTER> [,<FILTERCHAIN>]
211 <FILTERGRAPH> ::= [sws_flags=<flags>;] <FILTERCHAIN> [;<FILTERGRAPH>]
212 Notes on filtergraph escaping
214 Filtergraph description composition entails several levels of escaping.
215 See the "Quoting and escaping" section in the ffmpeg-utils(1) manual
216 for more information about the employed escaping procedure.
217 A first level escaping affects the content of each filter option value,
219 which may contain the special character ":" used to separate values, or
220 one of the escaping characters "\'".
221 A second level escaping affects the whole filter description, which may
223 contain the escaping characters "\'" or the special characters "[],;"
224 used by the filtergraph description.
225 Finally, when you specify a filtergraph on a shell commandline, you
227 need to perform a third level escaping for the shell special characters
228 contained within it.
229 For example, consider the following string to be embedded in the
231 drawtext filter description text value:
232 this is a 'string': may contain one, or more, special characters
234 This string contains the "'" special escaping character, and the ":"
236 special character, so it needs to be escaped in this way:
237 text=this is a \'string\'\: may contain one, or more, special characters
239 A second level of escaping is required when embedding the filter
241 description in a filtergraph description, in order to escape all the
242 filtergraph special characters. Thus the example above becomes:
243 drawtext=text=this is a \\\'string\\\'\\: may contain one\, or more\, special characters
245 (note that in addition to the "\'" escaping special characters, also
247 "," needs to be escaped).
248 Finally an additional level of escaping is needed when writing the
250 filtergraph description in a shell command, which depends on the
251 escaping rules of the adopted shell. For example, assuming that "\" is
252 special and needs to be escaped with another "\", the previous string
253 will finally result in:
254 -vf "drawtext=text=this is a \\\\\\'string\\\\\\'\\\\: may contain one\\, or more\\, special characters"
256
258 Some filters support a generic enable option. For the filters
259 supporting timeline editing, this option can be set to an expression
260 which is evaluated before sending a frame to the filter. If the
261 evaluation is non-zero, the filter will be enabled, otherwise the frame
262 will be sent unchanged to the next filter in the filtergraph.
263 The expression accepts the following values:
265 t timestamp expressed in seconds, NAN if the input timestamp is
267 unknown
268 n sequential number of the input frame, starting from 0
270 pos the position in the file of the input frame, NAN if unknown
272 w
274 h width and height of the input frame if video
275 Additionally, these filters support an enable command that can be used
277 to re-define the expression.
278 Like any other filtering option, the enable option follows the same
280 rules.
281 For example, to enable a blur filter (smartblur) from 10 seconds to 3
283 minutes, and a curves filter starting at 3 seconds:
284 smartblur = enable='between(t,10,3*60)',
286 curves = enable='gte(t,3)' : preset=cross_process
287 See "ffmpeg -filters" to view which filters have timeline support.
289
291 Some options can be changed during the operation of the filter using a
292 command. These options are marked 'T' on the output of ffmpeg -h
293 filter=<name of filter>. The name of the command is the name of the
294 option and the argument is the new value.
295
297 Some filters with several inputs support a common set of options.
298 These options can only be set by name, not with the short notation.
299 eof_action
301 The action to take when EOF is encountered on the secondary input;
302 it accepts one of the following values:
303 repeat
305 Repeat the last frame (the default).
306 endall
308 End both streams.
309 pass
311 Pass the main input through.
312 shortest
314 If set to 1, force the output to terminate when the shortest input
315 terminates. Default value is 0.
316 repeatlast
318 If set to 1, force the filter to extend the last frame of secondary
319 streams until the end of the primary stream. A value of 0 disables
320 this behavior. Default value is 1.
321 ts_sync_mode
323 How strictly to sync streams based on secondary input timestamps;
324 it accepts one of the following values:
325 default
327 Frame from secondary input with the nearest lower or equal
328 timestamp to the primary input frame.
329 nearest
331 Frame from secondary input with the absolute nearest timestamp
332 to the primary input frame.
333
335 When you configure your FFmpeg build, you can disable any of the
336 existing filters using "--disable-filters". The configure output will
337 show the audio filters included in your build.
338 Below is a description of the currently available audio filters.
340 acompressor
342 A compressor is mainly used to reduce the dynamic range of a signal.
343 Especially modern music is mostly compressed at a high ratio to improve
344 the overall loudness. It's done to get the highest attention of a
345 listener, "fatten" the sound and bring more "power" to the track. If a
346 signal is compressed too much it may sound dull or "dead" afterwards or
347 it may start to "pump" (which could be a powerful effect but can also
348 destroy a track completely). The right compression is the key to reach
349 a professional sound and is the high art of mixing and mastering.
350 Because of its complex settings it may take a long time to get the
351 right feeling for this kind of effect.
352 Compression is done by detecting the volume above a chosen level
354 "threshold" and dividing it by the factor set with "ratio". So if you
355 set the threshold to -12dB and your signal reaches -6dB a ratio of 2:1
356 will result in a signal at -9dB. Because an exact manipulation of the
357 signal would cause distortion of the waveform the reduction can be
358 levelled over the time. This is done by setting "Attack" and "Release".
359 "attack" determines how long the signal has to rise above the threshold
360 before any reduction will occur and "release" sets the time the signal
361 has to fall below the threshold to reduce the reduction again. Shorter
362 signals than the chosen attack time will be left untouched. The
363 overall reduction of the signal can be made up afterwards with the
364 "makeup" setting. So compressing the peaks of a signal about 6dB and
365 raising the makeup to this level results in a signal twice as loud than
366 the source. To gain a softer entry in the compression the "knee"
367 flattens the hard edge at the threshold in the range of the chosen
368 decibels.
369 The filter accepts the following options:
371 level_in
373 Set input gain. Default is 1. Range is between 0.015625 and 64.
374 mode
376 Set mode of compressor operation. Can be "upward" or "downward".
377 Default is "downward".
378 threshold
380 If a signal of stream rises above this level it will affect the
381 gain reduction. By default it is 0.125. Range is between
382 0.00097563 and 1.
383 ratio
385 Set a ratio by which the signal is reduced. 1:2 means that if the
386 level rose 4dB above the threshold, it will be only 2dB above after
387 the reduction. Default is 2. Range is between 1 and 20.
388 attack
390 Amount of milliseconds the signal has to rise above the threshold
391 before gain reduction starts. Default is 20. Range is between 0.01
392 and 2000.
393 release
395 Amount of milliseconds the signal has to fall below the threshold
396 before reduction is decreased again. Default is 250. Range is
397 between 0.01 and 9000.
398 makeup
400 Set the amount by how much signal will be amplified after
401 processing. Default is 1. Range is from 1 to 64.
402 knee
404 Curve the sharp knee around the threshold to enter gain reduction
405 more softly. Default is 2.82843. Range is between 1 and 8.
406 link
408 Choose if the "average" level between all channels of input stream
409 or the louder("maximum") channel of input stream affects the
410 reduction. Default is "average".
411 detection
413 Should the exact signal be taken in case of "peak" or an RMS one in
414 case of "rms". Default is "rms" which is mostly smoother.
415 mix How much to use compressed signal in output. Default is 1. Range
417 is between 0 and 1.
418 Commands
420 This filter supports the all above options as commands.
422 acontrast
424 Simple audio dynamic range compression/expansion filter.
425 The filter accepts the following options:
427 contrast
429 Set contrast. Default is 33. Allowed range is between 0 and 100.
430 acopy
432 Copy the input audio source unchanged to the output. This is mainly
433 useful for testing purposes.
434 acrossfade
436 Apply cross fade from one input audio stream to another input audio
437 stream. The cross fade is applied for specified duration near the end
438 of first stream.
439 The filter accepts the following options:
441 nb_samples, ns
443 Specify the number of samples for which the cross fade effect has
444 to last. At the end of the cross fade effect the first input audio
445 will be completely silent. Default is 44100.
446 duration, d
448 Specify the duration of the cross fade effect. See the Time
449 duration section in the ffmpeg-utils(1) manual for the accepted
450 syntax. By default the duration is determined by nb_samples. If
451 set this option is used instead of nb_samples.
452 overlap, o
454 Should first stream end overlap with second stream start. Default
455 is enabled.
456 curve1
458 Set curve for cross fade transition for first stream.
459 curve2
461 Set curve for cross fade transition for second stream.
462 For description of available curve types see afade filter
464 description.
465 Examples
467 • Cross fade from one input to another:
469 ffmpeg -i first.flac -i second.flac -filter_complex acrossfade=d=10:c1=exp:c2=exp output.flac
471 • Cross fade from one input to another but without overlapping:
473 ffmpeg -i first.flac -i second.flac -filter_complex acrossfade=d=10:o=0:c1=exp:c2=exp output.flac
475 acrossover
477 Split audio stream into several bands.
478 This filter splits audio stream into two or more frequency ranges.
480 Summing all streams back will give flat output.
481 The filter accepts the following options:
483 split
485 Set split frequencies. Those must be positive and increasing.
486 order
488 Set filter order for each band split. This controls filter roll-off
489 or steepness of filter transfer function. Available values are:
490 2nd 12 dB per octave.
492 4th 24 dB per octave.
494 6th 36 dB per octave.
496 8th 48 dB per octave.
498 10th
500 60 dB per octave.
501 12th
503 72 dB per octave.
504 14th
506 84 dB per octave.
507 16th
509 96 dB per octave.
510 18th
512 108 dB per octave.
513 20th
515 120 dB per octave.
516 Default is 4th.
518 level
520 Set input gain level. Allowed range is from 0 to 1. Default value
521 is 1.
522 gains
524 Set output gain for each band. Default value is 1 for all bands.
525 precision
527 Set which precision to use when processing samples.
528 auto
530 Auto pick internal sample format depending on other filters.
531 float
533 Always use single-floating point precision sample format.
534 double
536 Always use double-floating point precision sample format.
537 Default value is "auto".
539 Examples
541 • Split input audio stream into two bands (low and high) with split
543 frequency of 1500 Hz, each band will be in separate stream:
544 ffmpeg -i in.flac -filter_complex 'acrossover=split=1500[LOW][HIGH]' -map '[LOW]' low.wav -map '[HIGH]' high.wav
546 • Same as above, but with higher filter order:
548 ffmpeg -i in.flac -filter_complex 'acrossover=split=1500:order=8th[LOW][HIGH]' -map '[LOW]' low.wav -map '[HIGH]' high.wav
550 • Same as above, but also with additional middle band (frequencies
552 between 1500 and 8000):
553 ffmpeg -i in.flac -filter_complex 'acrossover=split=1500 8000:order=8th[LOW][MID][HIGH]' -map '[LOW]' low.wav -map '[MID]' mid.wav -map '[HIGH]' high.wav
555 acrusher
557 Reduce audio bit resolution.
558 This filter is bit crusher with enhanced functionality. A bit crusher
560 is used to audibly reduce number of bits an audio signal is sampled
561 with. This doesn't change the bit depth at all, it just produces the
562 effect. Material reduced in bit depth sounds more harsh and "digital".
563 This filter is able to even round to continuous values instead of
564 discrete bit depths. Additionally it has a D/C offset which results in
565 different crushing of the lower and the upper half of the signal. An
566 Anti-Aliasing setting is able to produce "softer" crushing sounds.
567 Another feature of this filter is the logarithmic mode. This setting
569 switches from linear distances between bits to logarithmic ones. The
570 result is a much more "natural" sounding crusher which doesn't gate low
571 signals for example. The human ear has a logarithmic perception, so
572 this kind of crushing is much more pleasant. Logarithmic crushing is
573 also able to get anti-aliased.
574 The filter accepts the following options:
576 level_in
578 Set level in.
579 level_out
581 Set level out.
582 bits
584 Set bit reduction.
585 mix Set mixing amount.
587 mode
589 Can be linear: "lin" or logarithmic: "log".
590 dc Set DC.
592 aa Set anti-aliasing.
594 samples
596 Set sample reduction.
597 lfo Enable LFO. By default disabled.
599 lforange
601 Set LFO range.
602 lforate
604 Set LFO rate.
605 Commands
607 This filter supports the all above options as commands.
609 acue
611 Delay audio filtering until a given wallclock timestamp. See the cue
612 filter.
613 adeclick
615 Remove impulsive noise from input audio.
616 Samples detected as impulsive noise are replaced by interpolated
618 samples using autoregressive modelling.
619 window, w
621 Set window size, in milliseconds. Allowed range is from 10 to 100.
622 Default value is 55 milliseconds. This sets size of window which
623 will be processed at once.
624 overlap, o
626 Set window overlap, in percentage of window size. Allowed range is
627 from 50 to 95. Default value is 75 percent. Setting this to a very
628 high value increases impulsive noise removal but makes whole
629 process much slower.
630 arorder, a
632 Set autoregression order, in percentage of window size. Allowed
633 range is from 0 to 25. Default value is 2 percent. This option also
634 controls quality of interpolated samples using neighbour good
635 samples.
636 threshold, t
638 Set threshold value. Allowed range is from 1 to 100. Default value
639 is 2. This controls the strength of impulsive noise which is going
640 to be removed. The lower value, the more samples will be detected
641 as impulsive noise.
642 burst, b
644 Set burst fusion, in percentage of window size. Allowed range is 0
645 to 10. Default value is 2. If any two samples detected as noise
646 are spaced less than this value then any sample between those two
647 samples will be also detected as noise.
648 method, m
650 Set overlap method.
651 It accepts the following values:
653 add, a
655 Select overlap-add method. Even not interpolated samples are
656 slightly changed with this method.
657 save, s
659 Select overlap-save method. Not interpolated samples remain
660 unchanged.
661 Default value is "a".
663 adeclip
665 Remove clipped samples from input audio.
666 Samples detected as clipped are replaced by interpolated samples using
668 autoregressive modelling.
669 window, w
671 Set window size, in milliseconds. Allowed range is from 10 to 100.
672 Default value is 55 milliseconds. This sets size of window which
673 will be processed at once.
674 overlap, o
676 Set window overlap, in percentage of window size. Allowed range is
677 from 50 to 95. Default value is 75 percent.
678 arorder, a
680 Set autoregression order, in percentage of window size. Allowed
681 range is from 0 to 25. Default value is 8 percent. This option also
682 controls quality of interpolated samples using neighbour good
683 samples.
684 threshold, t
686 Set threshold value. Allowed range is from 1 to 100. Default value
687 is 10. Higher values make clip detection less aggressive.
688 hsize, n
690 Set size of histogram used to detect clips. Allowed range is from
691 100 to 9999. Default value is 1000. Higher values make clip
692 detection less aggressive.
693 method, m
695 Set overlap method.
696 It accepts the following values:
698 add, a
700 Select overlap-add method. Even not interpolated samples are
701 slightly changed with this method.
702 save, s
704 Select overlap-save method. Not interpolated samples remain
705 unchanged.
706 Default value is "a".
708 adecorrelate
710 Apply decorrelation to input audio stream.
711 The filter accepts the following options:
713 stages
715 Set decorrelation stages of filtering. Allowed range is from 1 to
716 16. Default value is 6.
717 seed
719 Set random seed used for setting delay in samples across channels.
720 adelay
722 Delay one or more audio channels.
723 Samples in delayed channel are filled with silence.
725 The filter accepts the following option:
727 delays
729 Set list of delays in milliseconds for each channel separated by
730 '|'. Unused delays will be silently ignored. If number of given
731 delays is smaller than number of channels all remaining channels
732 will not be delayed. If you want to delay exact number of samples,
733 append 'S' to number. If you want instead to delay in seconds,
734 append 's' to number.
735 all Use last set delay for all remaining channels. By default is
737 disabled. This option if enabled changes how option "delays" is
738 interpreted.
739 Examples
741 • Delay first channel by 1.5 seconds, the third channel by 0.5
743 seconds and leave the second channel (and any other channels that
744 may be present) unchanged.
745 adelay=1500|0|500
747 • Delay second channel by 500 samples, the third channel by 700
749 samples and leave the first channel (and any other channels that
750 may be present) unchanged.
751 adelay=0|500S|700S
753 • Delay all channels by same number of samples:
755 adelay=delays=64S:all=1
757 adenorm
759 Remedy denormals in audio by adding extremely low-level noise.
760 This filter shall be placed before any filter that can produce
762 denormals.
763 A description of the accepted parameters follows.
765 level
767 Set level of added noise in dB. Default is -351. Allowed range is
768 from -451 to -90.
769 type
771 Set type of added noise.
772 dc Add DC signal.
774 ac Add AC signal.
776 square
778 Add square signal.
779 pulse
781 Add pulse signal.
782 Default is "dc".
784 Commands
786 This filter supports the all above options as commands.
788 aderivative, aintegral
790 Compute derivative/integral of audio stream.
791 Applying both filters one after another produces original audio.
793 adrc
795 Apply spectral dynamic range controller filter to input audio stream.
796 A description of the accepted options follows.
798 transfer
800 Set the transfer expression.
801 The expression can contain the following constants:
803 ch current channel number
805 sn current sample number
807 nb_channels
809 number of channels
810 t timestamp expressed in seconds
812 sr sample rate
814 p current frequency power value, in dB
816 f current frequency in Hz
818 Default value is "p".
820 attack
822 Set the attack in milliseconds. Default is 50 milliseconds.
823 Allowed range is from 1 to 1000 milliseconds.
824 release
826 Set the release in milliseconds. Default is 100 milliseconds.
827 Allowed range is from 5 to 2000 milliseconds.
828 channels
830 Set which channels to filter, by default "all" channels in audio
831 stream are filtered.
832 Commands
834 This filter supports the all above options as commands.
836 Examples
838 • Apply spectral compression to all frequencies with threshold of -50
840 dB and 1:6 ratio:
841 adrc=transfer='if(gt(p,-50),-50+(p-(-50))/6,p)':attack=50:release=100
843 • Similar to above but with 1:2 ratio and filtering only front center
845 channel:
846 adrc=transfer='if(gt(p,-50),-50+(p-(-50))/2,p)':attack=50:release=100:channels=FC
848 • Apply spectral noise gate to all frequencies with threshold of -85
850 dB and with short attack time and short release time:
851 adrc=transfer='if(lte(p,-85),p-800,p)':attack=1:release=5
853 • Apply spectral expansion to all frequencies with threshold of -10
855 dB and 1:2 ratio:
856 adrc=transfer='if(lt(p,-10),-10+(p-(-10))*2,p)':attack=50:release=100
858 • Apply limiter to max -60 dB to all frequencies, with attack of 2 ms
860 and release of 10 ms:
861 adrc=transfer='min(p,-60)':attack=2:release=10
863 adynamicequalizer
865 Apply dynamic equalization to input audio stream.
866 A description of the accepted options follows.
868 threshold
870 Set the detection threshold used to trigger equalization.
871 Threshold detection is using bandpass filter. Default value is 0.
872 Allowed range is from 0 to 100.
873 dfrequency
875 Set the detection frequency in Hz used for bandpass filter used to
876 trigger equalization. Default value is 1000 Hz. Allowed range is
877 between 2 and 1000000 Hz.
878 dqfactor
880 Set the detection resonance factor for bandpass filter used to
881 trigger equalization. Default value is 1. Allowed range is from
882 0.001 to 1000.
883 tfrequency
885 Set the target frequency of equalization filter. Default value is
886 1000 Hz. Allowed range is between 2 and 1000000 Hz.
887 tqfactor
889 Set the target resonance factor for target equalization filter.
890 Default value is 1. Allowed range is from 0.001 to 1000.
891 attack
893 Set the amount of milliseconds the signal from detection has to
894 rise above the detection threshold before equalization starts.
895 Default is 20. Allowed range is between 1 and 2000.
896 release
898 Set the amount of milliseconds the signal from detection has to
899 fall below the detection threshold before equalization ends.
900 Default is 200. Allowed range is between 1 and 2000.
901 ratio
903 Set the ratio by which the equalization gain is raised. Default is
904 1. Allowed range is between 0 and 30.
905 makeup
907 Set the makeup offset by which the equalization gain is raised.
908 Default is 0. Allowed range is between 0 and 100.
909 range
911 Set the max allowed cut/boost amount. Default is 50. Allowed range
912 is from 1 to 200.
913 mode
915 Set the mode of filter operation, can be one of the following:
916 listen
918 Output only isolated bandpass signal.
919 cut Cut frequencies above detection threshold.
921 boost
923 Boost frequencies bellow detection threshold.
924 Default mode is cut.
926 tftype
928 Set the type of target filter, can be one of the following:
929 bell
931 lowshelf
932 highshelf
933 Default type is bell.
935 direction
937 Set processing direction relative to threshold.
938 downward
940 Boost/Cut if threshold is higher/lower than detected volume.
941 upward
943 Boost/Cut if threshold is lower/higher than detected volume.
944 Default direction is downward.
946 auto
948 Automatically gather threshold from detection filter. By default is
949 disabled. This option is useful to detect threshold in certain
950 time frame of input audio stream, in such case option value is
951 changed at runtime.
952 Available values are:
954 disabled
956 Disable using automatically gathered threshold value.
957 off Stop picking threshold value.
959 on Start picking threshold value.
961 Commands
963 This filter supports the all above options as commands.
965 adynamicsmooth
967 Apply dynamic smoothing to input audio stream.
968 A description of the accepted options follows.
970 sensitivity
972 Set an amount of sensitivity to frequency fluctations. Default is
973 2. Allowed range is from 0 to 1e+06.
974 basefreq
976 Set a base frequency for smoothing. Default value is 22050.
977 Allowed range is from 2 to 1e+06.
978 Commands
980 This filter supports the all above options as commands.
982 aecho
984 Apply echoing to the input audio.
985 Echoes are reflected sound and can occur naturally amongst mountains
987 (and sometimes large buildings) when talking or shouting; digital echo
988 effects emulate this behaviour and are often used to help fill out the
989 sound of a single instrument or vocal. The time difference between the
990 original signal and the reflection is the "delay", and the loudness of
991 the reflected signal is the "decay". Multiple echoes can have
992 different delays and decays.
993 A description of the accepted parameters follows.
995 in_gain
997 Set input gain of reflected signal. Default is 0.6.
998 out_gain
1000 Set output gain of reflected signal. Default is 0.3.
1001 delays
1003 Set list of time intervals in milliseconds between original signal
1004 and reflections separated by '|'. Allowed range for each "delay" is
1005 "(0 - 90000.0]". Default is 1000.
1006 decays
1008 Set list of loudness of reflected signals separated by '|'.
1009 Allowed range for each "decay" is "(0 - 1.0]". Default is 0.5.
1010 Examples
1012 • Make it sound as if there are twice as many instruments as are
1014 actually playing:
1015 aecho=0.8:0.88:60:0.4
1017 • If delay is very short, then it sounds like a (metallic) robot
1019 playing music:
1020 aecho=0.8:0.88:6:0.4
1022 • A longer delay will sound like an open air concert in the
1024 mountains:
1025 aecho=0.8:0.9:1000:0.3
1027 • Same as above but with one more mountain:
1029 aecho=0.8:0.9:1000|1800:0.3|0.25
1031 aemphasis
1033 Audio emphasis filter creates or restores material directly taken from
1034 LPs or emphased CDs with different filter curves. E.g. to store music
1035 on vinyl the signal has to be altered by a filter first to even out the
1036 disadvantages of this recording medium. Once the material is played
1037 back the inverse filter has to be applied to restore the distortion of
1038 the frequency response.
1039 The filter accepts the following options:
1041 level_in
1043 Set input gain.
1044 level_out
1046 Set output gain.
1047 mode
1049 Set filter mode. For restoring material use "reproduction" mode,
1050 otherwise use "production" mode. Default is "reproduction" mode.
1051 type
1053 Set filter type. Selects medium. Can be one of the following:
1054 col select Columbia.
1056 emi select EMI.
1058 bsi select BSI (78RPM).
1060 riaa
1062 select RIAA.
1063 cd select Compact Disc (CD).
1065 50fm
1067 select 50µs (FM).
1068 75fm
1070 select 75µs (FM).
1071 50kf
1073 select 50µs (FM-KF).
1074 75kf
1076 select 75µs (FM-KF).
1077 Commands
1079 This filter supports the all above options as commands.
1081 aeval
1083 Modify an audio signal according to the specified expressions.
1084 This filter accepts one or more expressions (one for each channel),
1086 which are evaluated and used to modify a corresponding audio signal.
1087 It accepts the following parameters:
1089 exprs
1091 Set the '|'-separated expressions list for each separate channel.
1092 If the number of input channels is greater than the number of
1093 expressions, the last specified expression is used for the
1094 remaining output channels.
1095 channel_layout, c
1097 Set output channel layout. If not specified, the channel layout is
1098 specified by the number of expressions. If set to same, it will use
1099 by default the same input channel layout.
1100 Each expression in exprs can contain the following constants and
1102 functions:
1103 ch channel number of the current expression
1105 n number of the evaluated sample, starting from 0
1107 s sample rate
1109 t time of the evaluated sample expressed in seconds
1111 nb_in_channels
1113 nb_out_channels
1114 input and output number of channels
1115 val(CH)
1117 the value of input channel with number CH
1118 Note: this filter is slow. For faster processing you should use a
1120 dedicated filter.
1121 Examples
1123 • Half volume:
1125 aeval=val(ch)/2:c=same
1127 • Invert phase of the second channel:
1129 aeval=val(0)|-val(1)
1131 aexciter
1133 An exciter is used to produce high sound that is not present in the
1134 original signal. This is done by creating harmonic distortions of the
1135 signal which are restricted in range and added to the original signal.
1136 An Exciter raises the upper end of an audio signal without simply
1137 raising the higher frequencies like an equalizer would do to create a
1138 more "crisp" or "brilliant" sound.
1139 The filter accepts the following options:
1141 level_in
1143 Set input level prior processing of signal. Allowed range is from
1144 0 to 64. Default value is 1.
1145 level_out
1147 Set output level after processing of signal. Allowed range is from
1148 0 to 64. Default value is 1.
1149 amount
1151 Set the amount of harmonics added to original signal. Allowed
1152 range is from 0 to 64. Default value is 1.
1153 drive
1155 Set the amount of newly created harmonics. Allowed range is from
1156 0.1 to 10. Default value is 8.5.
1157 blend
1159 Set the octave of newly created harmonics. Allowed range is from
1160 -10 to 10. Default value is 0.
1161 freq
1163 Set the lower frequency limit of producing harmonics in Hz.
1164 Allowed range is from 2000 to 12000 Hz. Default is 7500 Hz.
1165 ceil
1167 Set the upper frequency limit of producing harmonics. Allowed
1168 range is from 9999 to 20000 Hz. If value is lower than 10000 Hz no
1169 limit is applied.
1170 listen
1172 Mute the original signal and output only added harmonics. By
1173 default is disabled.
1174 Commands
1176 This filter supports the all above options as commands.
1178 afade
1180 Apply fade-in/out effect to input audio.
1181 A description of the accepted parameters follows.
1183 type, t
1185 Specify the effect type, can be either "in" for fade-in, or "out"
1186 for a fade-out effect. Default is "in".
1187 start_sample, ss
1189 Specify the number of the start sample for starting to apply the
1190 fade effect. Default is 0.
1191 nb_samples, ns
1193 Specify the number of samples for which the fade effect has to
1194 last. At the end of the fade-in effect the output audio will have
1195 the same volume as the input audio, at the end of the fade-out
1196 transition the output audio will be silence. Default is 44100.
1197 start_time, st
1199 Specify the start time of the fade effect. Default is 0. The value
1200 must be specified as a time duration; see the Time duration section
1201 in the ffmpeg-utils(1) manual for the accepted syntax. If set this
1202 option is used instead of start_sample.
1203 duration, d
1205 Specify the duration of the fade effect. See the Time duration
1206 section in the ffmpeg-utils(1) manual for the accepted syntax. At
1207 the end of the fade-in effect the output audio will have the same
1208 volume as the input audio, at the end of the fade-out transition
1209 the output audio will be silence. By default the duration is
1210 determined by nb_samples. If set this option is used instead of
1211 nb_samples.
1212 curve
1214 Set curve for fade transition.
1215 It accepts the following values:
1217 tri select triangular, linear slope (default)
1219 qsin
1221 select quarter of sine wave
1222 hsin
1224 select half of sine wave
1225 esin
1227 select exponential sine wave
1228 log select logarithmic
1230 ipar
1232 select inverted parabola
1233 qua select quadratic
1235 cub select cubic
1237 squ select square root
1239 cbr select cubic root
1241 par select parabola
1243 exp select exponential
1245 iqsin
1247 select inverted quarter of sine wave
1248 ihsin
1250 select inverted half of sine wave
1251 dese
1253 select double-exponential seat
1254 desi
1256 select double-exponential sigmoid
1257 losi
1259 select logistic sigmoid
1260 sinc
1262 select sine cardinal function
1263 isinc
1265 select inverted sine cardinal function
1266 nofade
1268 no fade applied
1269 silence
1271 Set the initial gain for fade-in or final gain for fade-out.
1272 Default value is 0.0.
1273 unity
1275 Set the initial gain for fade-out or final gain for fade-in.
1276 Default value is 1.0.
1277 Commands
1279 This filter supports the all above options as commands.
1281 Examples
1283 • Fade in first 15 seconds of audio:
1285 afade=t=in:ss=0:d=15
1287 • Fade out last 25 seconds of a 900 seconds audio:
1289 afade=t=out:st=875:d=25
1291 afftdn
1293 Denoise audio samples with FFT.
1294 A description of the accepted parameters follows.
1296 noise_reduction, nr
1298 Set the noise reduction in dB, allowed range is 0.01 to 97.
1299 Default value is 12 dB.
1300 noise_floor, nf
1302 Set the noise floor in dB, allowed range is -80 to -20. Default
1303 value is -50 dB.
1304 noise_type, nt
1306 Set the noise type.
1307 It accepts the following values:
1309 white, w
1311 Select white noise.
1312 vinyl, v
1314 Select vinyl noise.
1315 shellac, s
1317 Select shellac noise.
1318 custom, c
1320 Select custom noise, defined in "bn" option.
1321 Default value is white noise.
1323 band_noise, bn
1325 Set custom band noise profile for every one of 15 bands. Bands are
1326 separated by ' ' or '|'.
1327 residual_floor, rf
1329 Set the residual floor in dB, allowed range is -80 to -20. Default
1330 value is -38 dB.
1331 track_noise, tn
1333 Enable noise floor tracking. By default is disabled. With this
1334 enabled, noise floor is automatically adjusted.
1335 track_residual, tr
1337 Enable residual tracking. By default is disabled.
1338 output_mode, om
1340 Set the output mode.
1341 It accepts the following values:
1343 input, i
1345 Pass input unchanged.
1346 output, o
1348 Pass noise filtered out.
1349 noise, n
1351 Pass only noise.
1352 Default value is output.
1354 adaptivity, ad
1356 Set the adaptivity factor, used how fast to adapt gains adjustments
1357 per each frequency bin. Value 0 enables instant adaptation, while
1358 higher values react much slower. Allowed range is from 0 to 1.
1359 Default value is 0.5.
1360 floor_offset, fo
1362 Set the noise floor offset factor. This option is used to adjust
1363 offset applied to measured noise floor. It is only effective when
1364 noise floor tracking is enabled. Allowed range is from -2.0 to
1365 2.0. Default value is 1.0.
1366 noise_link, nl
1368 Set the noise link used for multichannel audio.
1369 It accepts the following values:
1371 none
1373 Use unchanged channel's noise floor.
1374 min Use measured min noise floor of all channels.
1376 max Use measured max noise floor of all channels.
1378 average
1380 Use measured average noise floor of all channels.
1381 Default value is min.
1383 band_multiplier, bm
1385 Set the band multiplier factor, used how much to spread bands
1386 across frequency bins. Allowed range is from 0.2 to 5. Default
1387 value is 1.25.
1388 sample_noise, sn
1390 Toggle capturing and measurement of noise profile from input audio.
1391 It accepts the following values:
1393 start, begin
1395 Start sample noise capture.
1396 stop, end
1398 Stop sample noise capture and measure new noise band profile.
1399 Default value is "none".
1401 gain_smooth, gs
1403 Set gain smooth spatial radius, used to smooth gains applied to
1404 each frequency bin. Useful to reduce random music noise artefacts.
1405 Higher values increases smoothing of gains. Allowed range is from
1406 0 to 50. Default value is 0.
1407 Commands
1409 This filter supports the some above mentioned options as commands.
1411 Examples
1413 • Reduce white noise by 10dB, and use previously measured noise floor
1415 of -40dB:
1416 afftdn=nr=10:nf=-40
1418 • Reduce white noise by 10dB, also set initial noise floor to -80dB
1420 and enable automatic tracking of noise floor so noise floor will
1421 gradually change during processing:
1422 afftdn=nr=10:nf=-80:tn=1
1424 • Reduce noise by 20dB, using noise floor of -40dB and using commands
1426 to take noise profile of first 0.4 seconds of input audio:
1427 asendcmd=0.0 afftdn sn start,asendcmd=0.4 afftdn sn stop,afftdn=nr=20:nf=-40
1429 afftfilt
1431 Apply arbitrary expressions to samples in frequency domain.
1432 real
1434 Set frequency domain real expression for each separate channel
1435 separated by '|'. Default is "re". If the number of input channels
1436 is greater than the number of expressions, the last specified
1437 expression is used for the remaining output channels.
1438 imag
1440 Set frequency domain imaginary expression for each separate channel
1441 separated by '|'. Default is "im".
1442 Each expression in real and imag can contain the following
1444 constants and functions:
1445 sr sample rate
1447 b current frequency bin number
1449 nb number of available bins
1451 ch channel number of the current expression
1453 chs number of channels
1455 pts current frame pts
1457 re current real part of frequency bin of current channel
1459 im current imaginary part of frequency bin of current channel
1461 real(b, ch)
1463 Return the value of real part of frequency bin at location
1464 (bin,channel)
1465 imag(b, ch)
1467 Return the value of imaginary part of frequency bin at location
1468 (bin,channel)
1469 win_size
1471 Set window size. Allowed range is from 16 to 131072. Default is
1472 4096
1473 win_func
1475 Set window function.
1476 It accepts the following values:
1478 rect
1480 bartlett
1481 hann, hanning
1482 hamming
1483 blackman
1484 welch
1485 flattop
1486 bharris
1487 bnuttall
1488 bhann
1489 sine
1490 nuttall
1491 lanczos
1492 gauss
1493 tukey
1494 dolph
1495 cauchy
1496 parzen
1497 poisson
1498 bohman
1499 kaiser
1500 Default is "hann".
1502 overlap
1504 Set window overlap. If set to 1, the recommended overlap for
1505 selected window function will be picked. Default is 0.75.
1506 Examples
1508 • Leave almost only low frequencies in audio:
1510 afftfilt="'real=re * (1-clip((b/nb)*b,0,1))':imag='im * (1-clip((b/nb)*b,0,1))'"
1512 • Apply robotize effect:
1514 afftfilt="real='hypot(re,im)*sin(0)':imag='hypot(re,im)*cos(0)':win_size=512:overlap=0.75"
1516 • Apply whisper effect:
1518 afftfilt="real='hypot(re,im)*cos((random(0)*2-1)*2*3.14)':imag='hypot(re,im)*sin((random(1)*2-1)*2*3.14)':win_size=128:overlap=0.8"
1520 • Apply phase shift:
1522 afftfilt="real=re*cos(1)-im*sin(1):imag=re*sin(1)+im*cos(1)"
1524 afir
1526 Apply an arbitrary Finite Impulse Response filter.
1527 This filter is designed for applying long FIR filters, up to 60 seconds
1529 long.
1530 It can be used as component for digital crossover filters, room
1532 equalization, cross talk cancellation, wavefield synthesis,
1533 auralization, ambiophonics, ambisonics and spatialization.
1534 This filter uses the streams higher than first one as FIR coefficients.
1536 If the non-first stream holds a single channel, it will be used for all
1537 input channels in the first stream, otherwise the number of channels in
1538 the non-first stream must be same as the number of channels in the
1539 first stream.
1540 It accepts the following parameters:
1542 dry Set dry gain. This sets input gain.
1544 wet Set wet gain. This sets final output gain.
1546 length
1548 Set Impulse Response filter length. Default is 1, which means whole
1549 IR is processed.
1550 gtype
1552 Enable applying gain measured from power of IR.
1553 Set which approach to use for auto gain measurement.
1555 none
1557 Do not apply any gain.
1558 peak
1560 select peak gain, very conservative approach. This is default
1561 value.
1562 dc select DC gain, limited application.
1564 gn select gain to noise approach, this is most popular one.
1566 ac select AC gain.
1568 rms select RMS gain.
1570 irgain
1572 Set gain to be applied to IR coefficients before filtering.
1573 Allowed range is 0 to 1. This gain is applied after any gain
1574 applied with gtype option.
1575 irfmt
1577 Set format of IR stream. Can be "mono" or "input". Default is
1578 "input".
1579 maxir
1581 Set max allowed Impulse Response filter duration in seconds.
1582 Default is 30 seconds. Allowed range is 0.1 to 60 seconds.
1583 response
1585 Show IR frequency response, magnitude(magenta), phase(green) and
1586 group delay(yellow) in additional video stream. By default it is
1587 disabled.
1588 channel
1590 Set for which IR channel to display frequency response. By default
1591 is first channel displayed. This option is used only when response
1592 is enabled.
1593 size
1595 Set video stream size. This option is used only when response is
1596 enabled.
1597 rate
1599 Set video stream frame rate. This option is used only when response
1600 is enabled.
1601 minp
1603 Set minimal partition size used for convolution. Default is 8192.
1604 Allowed range is from 1 to 65536. Lower values decreases latency
1605 at cost of higher CPU usage.
1606 maxp
1608 Set maximal partition size used for convolution. Default is 8192.
1609 Allowed range is from 8 to 65536. Lower values may increase CPU
1610 usage.
1611 nbirs
1613 Set number of input impulse responses streams which will be
1614 switchable at runtime. Allowed range is from 1 to 32. Default is
1615 1.
1616 ir Set IR stream which will be used for convolution, starting from 0,
1618 should always be lower than supplied value by "nbirs" option.
1619 Default is 0. This option can be changed at runtime via commands.
1620 precision
1622 Set which precision to use when processing samples.
1623 auto
1625 Auto pick internal sample format depending on other filters.
1626 float
1628 Always use single-floating point precision sample format.
1629 double
1631 Always use double-floating point precision sample format.
1632 Default value is auto.
1634 Examples
1636 • Apply reverb to stream using mono IR file as second input, complete
1638 command using ffmpeg:
1639 ffmpeg -i input.wav -i middle_tunnel_1way_mono.wav -lavfi afir output.wav
1641 • Apply true stereo processing given input stereo stream, and two
1643 stereo impulse responses for left and right channel, the impulse
1644 response files are files with names l_ir.wav and r_ir.wav:
1645 "pan=4C|c0=FL|c1=FL|c2=FR|c3=FR[a];amovie=l_ir.wav[LIR];amovie=r_ir.wav[RIR];[LIR][RIR]amerge[ir];[a][ir]afir=irfmt=input:gtype=gn:irgain=-5dB,pan=stereo|FL<c0+c2|FR<c1+c3"
1647 aformat
1649 Set output format constraints for the input audio. The framework will
1650 negotiate the most appropriate format to minimize conversions.
1651 It accepts the following parameters:
1653 sample_fmts, f
1655 A '|'-separated list of requested sample formats.
1656 sample_rates, r
1658 A '|'-separated list of requested sample rates.
1659 channel_layouts, cl
1661 A '|'-separated list of requested channel layouts.
1662 See the Channel Layout section in the ffmpeg-utils(1) manual for
1664 the required syntax.
1665 If a parameter is omitted, all values are allowed.
1667 Force the output to either unsigned 8-bit or signed 16-bit stereo
1669 aformat=sample_fmts=u8|s16:channel_layouts=stereo
1671 afreqshift
1673 Apply frequency shift to input audio samples.
1674 The filter accepts the following options:
1676 shift
1678 Specify frequency shift. Allowed range is -INT_MAX to INT_MAX.
1679 Default value is 0.0.
1680 level
1682 Set output gain applied to final output. Allowed range is from 0.0
1683 to 1.0. Default value is 1.0.
1684 order
1686 Set filter order used for filtering. Allowed range is from 1 to 16.
1687 Default value is 8.
1688 Commands
1690 This filter supports the all above options as commands.
1692 afwtdn
1694 Reduce broadband noise from input samples using Wavelets.
1695 A description of the accepted options follows.
1697 sigma
1699 Set the noise sigma, allowed range is from 0 to 1. Default value
1700 is 0. This option controls strength of denoising applied to input
1701 samples. Most useful way to set this option is via decibels, eg.
1702 -45dB.
1703 levels
1705 Set the number of wavelet levels of decomposition. Allowed range
1706 is from 1 to 12. Default value is 10. Setting this too low make
1707 denoising performance very poor.
1708 wavet
1710 Set wavelet type for decomposition of input frame. They are sorted
1711 by number of coefficients, from lowest to highest. More
1712 coefficients means worse filtering speed, but overall better
1713 quality. Available wavelets are:
1714 sym2
1716 sym4
1717 rbior68
1718 deb10
1719 sym10
1720 coif5
1721 bl3
1722 percent
1723 Set percent of full denoising. Allowed range is from 0 to 100
1724 percent. Default value is 85 percent or partial denoising.
1725 profile
1727 If enabled, first input frame will be used as noise profile. If
1728 first frame samples contain non-noise performance will be very
1729 poor.
1730 adaptive
1732 If enabled, input frames are analyzed for presence of noise. If
1733 noise is detected with high possibility then input frame profile
1734 will be used for processing following frames, until new noise frame
1735 is detected.
1736 samples
1738 Set size of single frame in number of samples. Allowed range is
1739 from 512 to 65536. Default frame size is 8192 samples.
1740 softness
1742 Set softness applied inside thresholding function. Allowed range is
1743 from 0 to 10. Default softness is 1.
1744 Commands
1746 This filter supports the all above options as commands.
1748 agate
1750 A gate is mainly used to reduce lower parts of a signal. This kind of
1751 signal processing reduces disturbing noise between useful signals.
1752 Gating is done by detecting the volume below a chosen level threshold
1754 and dividing it by the factor set with ratio. The bottom of the noise
1755 floor is set via range. Because an exact manipulation of the signal
1756 would cause distortion of the waveform the reduction can be levelled
1757 over time. This is done by setting attack and release.
1758 attack determines how long the signal has to fall below the threshold
1760 before any reduction will occur and release sets the time the signal
1761 has to rise above the threshold to reduce the reduction again. Shorter
1762 signals than the chosen attack time will be left untouched.
1763 level_in
1765 Set input level before filtering. Default is 1. Allowed range is
1766 from 0.015625 to 64.
1767 mode
1769 Set the mode of operation. Can be "upward" or "downward". Default
1770 is "downward". If set to "upward" mode, higher parts of signal will
1771 be amplified, expanding dynamic range in upward direction.
1772 Otherwise, in case of "downward" lower parts of signal will be
1773 reduced.
1774 range
1776 Set the level of gain reduction when the signal is below the
1777 threshold. Default is 0.06125. Allowed range is from 0 to 1.
1778 Setting this to 0 disables reduction and then filter behaves like
1779 expander.
1780 threshold
1782 If a signal rises above this level the gain reduction is released.
1783 Default is 0.125. Allowed range is from 0 to 1.
1784 ratio
1786 Set a ratio by which the signal is reduced. Default is 2. Allowed
1787 range is from 1 to 9000.
1788 attack
1790 Amount of milliseconds the signal has to rise above the threshold
1791 before gain reduction stops. Default is 20 milliseconds. Allowed
1792 range is from 0.01 to 9000.
1793 release
1795 Amount of milliseconds the signal has to fall below the threshold
1796 before the reduction is increased again. Default is 250
1797 milliseconds. Allowed range is from 0.01 to 9000.
1798 makeup
1800 Set amount of amplification of signal after processing. Default is
1801 1. Allowed range is from 1 to 64.
1802 knee
1804 Curve the sharp knee around the threshold to enter gain reduction
1805 more softly. Default is 2.828427125. Allowed range is from 1 to 8.
1806 detection
1808 Choose if exact signal should be taken for detection or an RMS like
1809 one. Default is "rms". Can be "peak" or "rms".
1810 link
1812 Choose if the average level between all channels or the louder
1813 channel affects the reduction. Default is "average". Can be
1814 "average" or "maximum".
1815 Commands
1817 This filter supports the all above options as commands.
1819 aiir
1821 Apply an arbitrary Infinite Impulse Response filter.
1822 It accepts the following parameters:
1824 zeros, z
1826 Set B/numerator/zeros/reflection coefficients.
1827 poles, p
1829 Set A/denominator/poles/ladder coefficients.
1830 gains, k
1832 Set channels gains.
1833 dry_gain
1835 Set input gain.
1836 wet_gain
1838 Set output gain.
1839 format, f
1841 Set coefficients format.
1842 ll lattice-ladder function
1844 sf analog transfer function
1846 tf digital transfer function
1848 zp Z-plane zeros/poles, cartesian (default)
1850 pr Z-plane zeros/poles, polar radians
1852 pd Z-plane zeros/poles, polar degrees
1854 sp S-plane zeros/poles
1856 process, r
1858 Set type of processing.
1859 d direct processing
1861 s serial processing
1863 p parallel processing
1865 precision, e
1867 Set filtering precision.
1868 dbl double-precision floating-point (default)
1870 flt single-precision floating-point
1872 i32 32-bit integers
1874 i16 16-bit integers
1876 normalize, n
1878 Normalize filter coefficients, by default is enabled. Enabling it
1879 will normalize magnitude response at DC to 0dB.
1880 mix How much to use filtered signal in output. Default is 1. Range is
1882 between 0 and 1.
1883 response
1885 Show IR frequency response, magnitude(magenta), phase(green) and
1886 group delay(yellow) in additional video stream. By default it is
1887 disabled.
1888 channel
1890 Set for which IR channel to display frequency response. By default
1891 is first channel displayed. This option is used only when response
1892 is enabled.
1893 size
1895 Set video stream size. This option is used only when response is
1896 enabled.
1897 Coefficients in "tf" and "sf" format are separated by spaces and are in
1899 ascending order.
1900 Coefficients in "zp" format are separated by spaces and order of
1902 coefficients doesn't matter. Coefficients in "zp" format are complex
1903 numbers with i imaginary unit.
1904 Different coefficients and gains can be provided for every channel, in
1906 such case use '|' to separate coefficients or gains. Last provided
1907 coefficients will be used for all remaining channels.
1908 Examples
1910 • Apply 2 pole elliptic notch at around 5000Hz for 48000 Hz sample
1912 rate:
1913 aiir=k=1:z=7.957584807809675810E-1 -2.575128568908332300 3.674839853930788710 -2.57512875289799137 7.957586296317130880E-1:p=1 -2.86950072432325953 3.63022088054647218 -2.28075678147272232 6.361362326477423500E-1:f=tf:r=d
1915 • Same as above but in "zp" format:
1917 aiir=k=0.79575848078096756:z=0.80918701+0.58773007i 0.80918701-0.58773007i 0.80884700+0.58784055i 0.80884700-0.58784055i:p=0.63892345+0.59951235i 0.63892345-0.59951235i 0.79582691+0.44198673i 0.79582691-0.44198673i:f=zp:r=s
1919 • Apply 3-rd order analog normalized Butterworth low-pass filter,
1921 using analog transfer function format:
1922 aiir=z=1.3057 0 0 0:p=1.3057 2.3892 2.1860 1:f=sf:r=d
1924 alimiter
1926 The limiter prevents an input signal from rising over a desired
1927 threshold. This limiter uses lookahead technology to prevent your
1928 signal from distorting. It means that there is a small delay after the
1929 signal is processed. Keep in mind that the delay it produces is the
1930 attack time you set.
1931 The filter accepts the following options:
1933 level_in
1935 Set input gain. Default is 1.
1936 level_out
1938 Set output gain. Default is 1.
1939 limit
1941 Don't let signals above this level pass the limiter. Default is 1.
1942 attack
1944 The limiter will reach its attenuation level in this amount of time
1945 in milliseconds. Default is 5 milliseconds.
1946 release
1948 Come back from limiting to attenuation 1.0 in this amount of
1949 milliseconds. Default is 50 milliseconds.
1950 asc When gain reduction is always needed ASC takes care of releasing to
1952 an average reduction level rather than reaching a reduction of 0 in
1953 the release time.
1954 asc_level
1956 Select how much the release time is affected by ASC, 0 means nearly
1957 no changes in release time while 1 produces higher release times.
1958 level
1960 Auto level output signal. Default is enabled. This normalizes
1961 audio back to 0dB if enabled.
1962 latency
1964 Compensate the delay introduced by using the lookahead buffer set
1965 with attack parameter. Also flush the valid audio data in the
1966 lookahead buffer when the stream hits EOF.
1967 Depending on picked setting it is recommended to upsample input 2x or
1969 4x times with aresample before applying this filter.
1970 allpass
1972 Apply a two-pole all-pass filter with central frequency (in Hz)
1973 frequency, and filter-width width. An all-pass filter changes the
1974 audio's frequency to phase relationship without changing its frequency
1975 to amplitude relationship.
1976 The filter accepts the following options:
1978 frequency, f
1980 Set frequency in Hz.
1981 width_type, t
1983 Set method to specify band-width of filter.
1984 h Hz
1986 q Q-Factor
1988 o octave
1990 s slope
1992 k kHz
1994 width, w
1996 Specify the band-width of a filter in width_type units.
1997 mix, m
1999 How much to use filtered signal in output. Default is 1. Range is
2000 between 0 and 1.
2001 channels, c
2003 Specify which channels to filter, by default all available are
2004 filtered.
2005 normalize, n
2007 Normalize biquad coefficients, by default is disabled. Enabling it
2008 will normalize magnitude response at DC to 0dB.
2009 order, o
2011 Set the filter order, can be 1 or 2. Default is 2.
2012 transform, a
2014 Set transform type of IIR filter.
2015 di
2017 dii
2018 tdi
2019 tdii
2020 latt
2021 svf
2022 zdf
2023 precision, r
2024 Set precison of filtering.
2025 auto
2027 Pick automatic sample format depending on surround filters.
2028 s16 Always use signed 16-bit.
2030 s32 Always use signed 32-bit.
2032 f32 Always use float 32-bit.
2034 f64 Always use float 64-bit.
2036 Commands
2038 This filter supports the following commands:
2040 frequency, f
2042 Change allpass frequency. Syntax for the command is : "frequency"
2043 width_type, t
2045 Change allpass width_type. Syntax for the command is :
2046 "width_type"
2047 width, w
2049 Change allpass width. Syntax for the command is : "width"
2050 mix, m
2052 Change allpass mix. Syntax for the command is : "mix"
2053 aloop
2055 Loop audio samples.
2056 The filter accepts the following options:
2058 loop
2060 Set the number of loops. Setting this value to -1 will result in
2061 infinite loops. Default is 0.
2062 size
2064 Set maximal number of samples. Default is 0.
2065 start
2067 Set first sample of loop. Default is 0.
2068 amerge
2070 Merge two or more audio streams into a single multi-channel stream.
2071 The filter accepts the following options:
2073 inputs
2075 Set the number of inputs. Default is 2.
2076 If the channel layouts of the inputs are disjoint, and therefore
2078 compatible, the channel layout of the output will be set accordingly
2079 and the channels will be reordered as necessary. If the channel layouts
2080 of the inputs are not disjoint, the output will have all the channels
2081 of the first input then all the channels of the second input, in that
2082 order, and the channel layout of the output will be the default value
2083 corresponding to the total number of channels.
2084 For example, if the first input is in 2.1 (FL+FR+LF) and the second
2086 input is FC+BL+BR, then the output will be in 5.1, with the channels in
2087 the following order: a1, a2, b1, a3, b2, b3 (a1 is the first channel of
2088 the first input, b1 is the first channel of the second input).
2089 On the other hand, if both input are in stereo, the output channels
2091 will be in the default order: a1, a2, b1, b2, and the channel layout
2092 will be arbitrarily set to 4.0, which may or may not be the expected
2093 value.
2094 All inputs must have the same sample rate, and format.
2096 If inputs do not have the same duration, the output will stop with the
2098 shortest.
2099 Examples
2101 • Merge two mono files into a stereo stream:
2103 amovie=left.wav [l] ; amovie=right.mp3 [r] ; [l] [r] amerge
2105 • Multiple merges assuming 1 video stream and 6 audio streams in
2107 input.mkv:
2108 ffmpeg -i input.mkv -filter_complex "[0:1][0:2][0:3][0:4][0:5][0:6] amerge=inputs=6" -c:a pcm_s16le output.mkv
2110 amix
2112 Mixes multiple audio inputs into a single output.
2113 Note that this filter only supports float samples (the amerge and pan
2115 audio filters support many formats). If the amix input has integer
2116 samples then aresample will be automatically inserted to perform the
2117 conversion to float samples.
2118 It accepts the following parameters:
2120 inputs
2122 The number of inputs. If unspecified, it defaults to 2.
2123 duration
2125 How to determine the end-of-stream.
2126 longest
2128 The duration of the longest input. (default)
2129 shortest
2131 The duration of the shortest input.
2132 first
2134 The duration of the first input.
2135 dropout_transition
2137 The transition time, in seconds, for volume renormalization when an
2138 input stream ends. The default value is 2 seconds.
2139 weights
2141 Specify weight of each input audio stream as a sequence of numbers
2142 separated by a space. If fewer weights are specified compared to
2143 number of inputs, the last weight is assigned to the remaining
2144 inputs. Default weight for each input is 1.
2145 normalize
2147 Always scale inputs instead of only doing summation of samples.
2148 Beware of heavy clipping if inputs are not normalized prior or
2149 after filtering by this filter if this option is disabled. By
2150 default is enabled.
2151 Examples
2153 • This will mix 3 input audio streams to a single output with the
2155 same duration as the first input and a dropout transition time of 3
2156 seconds:
2157 ffmpeg -i INPUT1 -i INPUT2 -i INPUT3 -filter_complex amix=inputs=3:duration=first:dropout_transition=3 OUTPUT
2159 • This will mix one vocal and one music input audio stream to a
2161 single output with the same duration as the longest input. The
2162 music will have quarter the weight as the vocals, and the inputs
2163 are not normalized:
2164 ffmpeg -i VOCALS -i MUSIC -filter_complex amix=inputs=2:duration=longest:dropout_transition=0:weights="1 0.25":normalize=0 OUTPUT
2166 Commands
2168 This filter supports the following commands:
2170 weights
2172 normalize
2173 Syntax is same as option with same name.
2174 amultiply
2176 Multiply first audio stream with second audio stream and store result
2177 in output audio stream. Multiplication is done by multiplying each
2178 sample from first stream with sample at same position from second
2179 stream.
2180 With this element-wise multiplication one can create amplitude fades
2182 and amplitude modulations.
2183 anequalizer
2185 High-order parametric multiband equalizer for each channel.
2186 It accepts the following parameters:
2188 params
2190 This option string is in format: "cchn f=cf w=w g=g t=f | ..."
2191 Each equalizer band is separated by '|'.
2192 chn Set channel number to which equalization will be applied. If
2194 input doesn't have that channel the entry is ignored.
2195 f Set central frequency for band. If input doesn't have that
2197 frequency the entry is ignored.
2198 w Set band width in Hertz.
2200 g Set band gain in dB.
2202 t Set filter type for band, optional, can be:
2204 0 Butterworth, this is default.
2206 1 Chebyshev type 1.
2208 2 Chebyshev type 2.
2210 curves
2212 With this option activated frequency response of anequalizer is
2213 displayed in video stream.
2214 size
2216 Set video stream size. Only useful if curves option is activated.
2217 mgain
2219 Set max gain that will be displayed. Only useful if curves option
2220 is activated. Setting this to a reasonable value makes it possible
2221 to display gain which is derived from neighbour bands which are too
2222 close to each other and thus produce higher gain when both are
2223 activated.
2224 fscale
2226 Set frequency scale used to draw frequency response in video
2227 output. Can be linear or logarithmic. Default is logarithmic.
2228 colors
2230 Set color for each channel curve which is going to be displayed in
2231 video stream. This is list of color names separated by space or by
2232 '|'. Unrecognised or missing colors will be replaced by white
2233 color.
2234 Examples
2236 • Lower gain by 10 of central frequency 200Hz and width 100 Hz for
2238 first 2 channels using Chebyshev type 1 filter:
2239 anequalizer=c0 f=200 w=100 g=-10 t=1|c1 f=200 w=100 g=-10 t=1
2241 Commands
2243 This filter supports the following commands:
2245 change
2247 Alter existing filter parameters. Syntax for the commands is :
2248 "fN|f=freq|w=width|g=gain"
2249 fN is existing filter number, starting from 0, if no such filter is
2251 available error is returned. freq set new frequency parameter.
2252 width set new width parameter in Hertz. gain set new gain
2253 parameter in dB.
2254 Full filter invocation with asendcmd may look like this:
2256 asendcmd=c='4.0 anequalizer change
2257 0|f=200|w=50|g=1',anequalizer=...
2258 anlmdn
2260 Reduce broadband noise in audio samples using Non-Local Means
2261 algorithm.
2262 Each sample is adjusted by looking for other samples with similar
2264 contexts. This context similarity is defined by comparing their
2265 surrounding patches of size p. Patches are searched in an area of r
2266 around the sample.
2267 The filter accepts the following options:
2269 strength, s
2271 Set denoising strength. Allowed range is from 0.00001 to 10000.
2272 Default value is 0.00001.
2273 patch, p
2275 Set patch radius duration. Allowed range is from 1 to 100
2276 milliseconds. Default value is 2 milliseconds.
2277 research, r
2279 Set research radius duration. Allowed range is from 2 to 300
2280 milliseconds. Default value is 6 milliseconds.
2281 output, o
2283 Set the output mode.
2284 It accepts the following values:
2286 i Pass input unchanged.
2288 o Pass noise filtered out.
2290 n Pass only noise.
2292 Default value is o.
2294 smooth, m
2296 Set smooth factor. Default value is 11. Allowed range is from 1 to
2297 1000.
2298 Commands
2300 This filter supports the all above options as commands.
2302 anlmf, anlms
2304 Apply Normalized Least-Mean-(Squares|Fourth) algorithm to the first
2305 audio stream using the second audio stream.
2306 This adaptive filter is used to mimic a desired filter by finding the
2308 filter coefficients that relate to producing the least mean square of
2309 the error signal (difference between the desired, 2nd input audio
2310 stream and the actual signal, the 1st input audio stream).
2311 A description of the accepted options follows.
2313 order
2315 Set filter order.
2316 mu Set filter mu.
2318 eps Set the filter eps.
2320 leakage
2322 Set the filter leakage.
2323 out_mode
2325 It accepts the following values:
2326 i Pass the 1st input.
2328 d Pass the 2nd input.
2330 o Pass filtered samples.
2332 n Pass difference between desired and filtered samples.
2334 Default value is o.
2336 Examples
2338 • One of many usages of this filter is noise reduction, input audio
2340 is filtered with same samples that are delayed by fixed amount, one
2341 such example for stereo audio is:
2342 asplit[a][b],[a]adelay=32S|32S[a],[b][a]anlms=order=128:leakage=0.0005:mu=.5:out_mode=o
2344 Commands
2346 This filter supports the same commands as options, excluding option
2348 "order".
2349 anull
2351 Pass the audio source unchanged to the output.
2352 apad
2354 Pad the end of an audio stream with silence.
2355 This can be used together with ffmpeg -shortest to extend audio streams
2357 to the same length as the video stream.
2358 A description of the accepted options follows.
2360 packet_size
2362 Set silence packet size. Default value is 4096.
2363 pad_len
2365 Set the number of samples of silence to add to the end. After the
2366 value is reached, the stream is terminated. This option is mutually
2367 exclusive with whole_len.
2368 whole_len
2370 Set the minimum total number of samples in the output audio stream.
2371 If the value is longer than the input audio length, silence is
2372 added to the end, until the value is reached. This option is
2373 mutually exclusive with pad_len.
2374 pad_dur
2376 Specify the duration of samples of silence to add. See the Time
2377 duration section in the ffmpeg-utils(1) manual for the accepted
2378 syntax. Used only if set to non-negative value.
2379 whole_dur
2381 Specify the minimum total duration in the output audio stream. See
2382 the Time duration section in the ffmpeg-utils(1) manual for the
2383 accepted syntax. Used only if set to non-negative value. If the
2384 value is longer than the input audio length, silence is added to
2385 the end, until the value is reached. This option is mutually
2386 exclusive with pad_dur
2387 If neither the pad_len nor the whole_len nor pad_dur nor whole_dur
2389 option is set, the filter will add silence to the end of the input
2390 stream indefinitely.
2391 Note that for ffmpeg 4.4 and earlier a zero pad_dur or whole_dur also
2393 caused the filter to add silence indefinitely.
2394 Examples
2396 • Add 1024 samples of silence to the end of the input:
2398 apad=pad_len=1024
2400 • Make sure the audio output will contain at least 10000 samples, pad
2402 the input with silence if required:
2403 apad=whole_len=10000
2405 • Use ffmpeg to pad the audio input with silence, so that the video
2407 stream will always result the shortest and will be converted until
2408 the end in the output file when using the shortest option:
2409 ffmpeg -i VIDEO -i AUDIO -filter_complex "[1:0]apad" -shortest OUTPUT
2411 aphaser
2413 Add a phasing effect to the input audio.
2414 A phaser filter creates series of peaks and troughs in the frequency
2416 spectrum. The position of the peaks and troughs are modulated so that
2417 they vary over time, creating a sweeping effect.
2418 A description of the accepted parameters follows.
2420 in_gain
2422 Set input gain. Default is 0.4.
2423 out_gain
2425 Set output gain. Default is 0.74
2426 delay
2428 Set delay in milliseconds. Default is 3.0.
2429 decay
2431 Set decay. Default is 0.4.
2432 speed
2434 Set modulation speed in Hz. Default is 0.5.
2435 type
2437 Set modulation type. Default is triangular.
2438 It accepts the following values:
2440 triangular, t
2442 sinusoidal, s
2443 aphaseshift
2445 Apply phase shift to input audio samples.
2446 The filter accepts the following options:
2448 shift
2450 Specify phase shift. Allowed range is from -1.0 to 1.0. Default
2451 value is 0.0.
2452 level
2454 Set output gain applied to final output. Allowed range is from 0.0
2455 to 1.0. Default value is 1.0.
2456 order
2458 Set filter order used for filtering. Allowed range is from 1 to 16.
2459 Default value is 8.
2460 Commands
2462 This filter supports the all above options as commands.
2464 apsyclip
2466 Apply Psychoacoustic clipper to input audio stream.
2467 The filter accepts the following options:
2469 level_in
2471 Set input gain. By default it is 1. Range is [0.015625 - 64].
2472 level_out
2474 Set output gain. By default it is 1. Range is [0.015625 - 64].
2475 clip
2477 Set the clipping start value. Default value is 0dBFS or 1.
2478 diff
2480 Output only difference samples, useful to hear introduced
2481 distortions. By default is disabled.
2482 adaptive
2484 Set strength of adaptive distortion applied. Default value is 0.5.
2485 Allowed range is from 0 to 1.
2486 iterations
2488 Set number of iterations of psychoacoustic clipper. Allowed range
2489 is from 1 to 20. Default value is 10.
2490 level
2492 Auto level output signal. Default is disabled. This normalizes
2493 audio back to 0dBFS if enabled.
2494 Commands
2496 This filter supports the all above options as commands.
2498 apulsator
2500 Audio pulsator is something between an autopanner and a tremolo. But
2501 it can produce funny stereo effects as well. Pulsator changes the
2502 volume of the left and right channel based on a LFO (low frequency
2503 oscillator) with different waveforms and shifted phases. This filter
2504 have the ability to define an offset between left and right channel. An
2505 offset of 0 means that both LFO shapes match each other. The left and
2506 right channel are altered equally - a conventional tremolo. An offset
2507 of 50% means that the shape of the right channel is exactly shifted in
2508 phase (or moved backwards about half of the frequency) - pulsator acts
2509 as an autopanner. At 1 both curves match again. Every setting in
2510 between moves the phase shift gapless between all stages and produces
2511 some "bypassing" sounds with sine and triangle waveforms. The more you
2512 set the offset near 1 (starting from the 0.5) the faster the signal
2513 passes from the left to the right speaker.
2514 The filter accepts the following options:
2516 level_in
2518 Set input gain. By default it is 1. Range is [0.015625 - 64].
2519 level_out
2521 Set output gain. By default it is 1. Range is [0.015625 - 64].
2522 mode
2524 Set waveform shape the LFO will use. Can be one of: sine, triangle,
2525 square, sawup or sawdown. Default is sine.
2526 amount
2528 Set modulation. Define how much of original signal is affected by
2529 the LFO.
2530 offset_l
2532 Set left channel offset. Default is 0. Allowed range is [0 - 1].
2533 offset_r
2535 Set right channel offset. Default is 0.5. Allowed range is [0 - 1].
2536 width
2538 Set pulse width. Default is 1. Allowed range is [0 - 2].
2539 timing
2541 Set possible timing mode. Can be one of: bpm, ms or hz. Default is
2542 hz.
2543 bpm Set bpm. Default is 120. Allowed range is [30 - 300]. Only used if
2545 timing is set to bpm.
2546 ms Set ms. Default is 500. Allowed range is [10 - 2000]. Only used if
2548 timing is set to ms.
2549 hz Set frequency in Hz. Default is 2. Allowed range is [0.01 - 100].
2551 Only used if timing is set to hz.
2552 aresample
2554 Resample the input audio to the specified parameters, using the
2555 libswresample library. If none are specified then the filter will
2556 automatically convert between its input and output.
2557 This filter is also able to stretch/squeeze the audio data to make it
2559 match the timestamps or to inject silence / cut out audio to make it
2560 match the timestamps, do a combination of both or do neither.
2561 The filter accepts the syntax [sample_rate:]resampler_options, where
2563 sample_rate expresses a sample rate and resampler_options is a list of
2564 key=value pairs, separated by ":". See the "Resampler Options" section
2565 in the ffmpeg-resampler(1) manual for the complete list of supported
2566 options.
2567 Examples
2569 • Resample the input audio to 44100Hz:
2571 aresample=44100
2573 • Stretch/squeeze samples to the given timestamps, with a maximum of
2575 1000 samples per second compensation:
2576 aresample=async=1000
2578 areverse
2580 Reverse an audio clip.
2581 Warning: This filter requires memory to buffer the entire clip, so
2583 trimming is suggested.
2584 Examples
2586 • Take the first 5 seconds of a clip, and reverse it.
2588 atrim=end=5,areverse
2590 arnndn
2592 Reduce noise from speech using Recurrent Neural Networks.
2593 This filter accepts the following options:
2595 model, m
2597 Set train model file to load. This option is always required.
2598 mix Set how much to mix filtered samples into final output. Allowed
2600 range is from -1 to 1. Default value is 1. Negative values are
2601 special, they set how much to keep filtered noise in the final
2602 filter output. Set this option to -1 to hear actual noise removed
2603 from input signal.
2604 Commands
2606 This filter supports the all above options as commands.
2608 asdr
2610 Measure Audio Signal-to-Distortion Ratio.
2611 This filter takes two audio streams for input, and outputs first audio
2613 stream. Results are in dB per channel at end of either input.
2614 asetnsamples
2616 Set the number of samples per each output audio frame.
2617 The last output packet may contain a different number of samples, as
2619 the filter will flush all the remaining samples when the input audio
2620 signals its end.
2621 The filter accepts the following options:
2623 nb_out_samples, n
2625 Set the number of frames per each output audio frame. The number is
2626 intended as the number of samples per each channel. Default value
2627 is 1024.
2628 pad, p
2630 If set to 1, the filter will pad the last audio frame with zeroes,
2631 so that the last frame will contain the same number of samples as
2632 the previous ones. Default value is 1.
2633 For example, to set the number of per-frame samples to 1234 and disable
2635 padding for the last frame, use:
2636 asetnsamples=n=1234:p=0
2638 asetrate
2640 Set the sample rate without altering the PCM data. This will result in
2641 a change of speed and pitch.
2642 The filter accepts the following options:
2644 sample_rate, r
2646 Set the output sample rate. Default is 44100 Hz.
2647 ashowinfo
2649 Show a line containing various information for each input audio frame.
2650 The input audio is not modified.
2651 The shown line contains a sequence of key/value pairs of the form
2653 key:value.
2654 The following values are shown in the output:
2656 n The (sequential) number of the input frame, starting from 0.
2658 pts The presentation timestamp of the input frame, in time base units;
2660 the time base depends on the filter input pad, and is usually
2661 1/sample_rate.
2662 pts_time
2664 The presentation timestamp of the input frame in seconds.
2665 pos position of the frame in the input stream, -1 if this information
2667 in unavailable and/or meaningless (for example in case of synthetic
2668 audio)
2669 fmt The sample format.
2671 chlayout
2673 The channel layout.
2674 rate
2676 The sample rate for the audio frame.
2677 nb_samples
2679 The number of samples (per channel) in the frame.
2680 checksum
2682 The Adler-32 checksum (printed in hexadecimal) of the audio data.
2683 For planar audio, the data is treated as if all the planes were
2684 concatenated.
2685 plane_checksums
2687 A list of Adler-32 checksums for each data plane.
2688 asoftclip
2690 Apply audio soft clipping.
2691 Soft clipping is a type of distortion effect where the amplitude of a
2693 signal is saturated along a smooth curve, rather than the abrupt shape
2694 of hard-clipping.
2695 This filter accepts the following options:
2697 type
2699 Set type of soft-clipping.
2700 It accepts the following values:
2702 hard
2704 tanh
2705 atan
2706 cubic
2707 exp
2708 alg
2709 quintic
2710 sin
2711 erf
2712 threshold
2713 Set threshold from where to start clipping. Default value is 0dB or
2714 1.
2715 output
2717 Set gain applied to output. Default value is 0dB or 1.
2718 param
2720 Set additional parameter which controls sigmoid function.
2721 oversample
2723 Set oversampling factor.
2724 Commands
2726 This filter supports the all above options as commands.
2728 aspectralstats
2730 Display frequency domain statistical information about the audio
2731 channels. Statistics are calculated and stored as metadata for each
2732 audio channel and for each audio frame.
2733 It accepts the following option:
2735 win_size
2737 Set the window length in samples. Default value is 2048. Allowed
2738 range is from 32 to 65536.
2739 win_func
2741 Set window function.
2742 It accepts the following values:
2744 rect
2746 bartlett
2747 hann, hanning
2748 hamming
2749 blackman
2750 welch
2751 flattop
2752 bharris
2753 bnuttall
2754 bhann
2755 sine
2756 nuttall
2757 lanczos
2758 gauss
2759 tukey
2760 dolph
2761 cauchy
2762 parzen
2763 poisson
2764 bohman
2765 kaiser
2766 Default is "hann".
2768 overlap
2770 Set window overlap. Allowed range is from 0 to 1. Default value is
2771 0.5.
2772 measure
2774 Select the parameters which are measured. The metadata keys can be
2775 used as flags, default is all which measures everything. none
2776 disables all measurement.
2777 A list of each metadata key follows:
2779 mean
2781 variance
2782 centroid
2783 spread
2784 skewness
2785 kurtosis
2786 entropy
2787 flatness
2788 crest
2789 flux
2790 slope
2791 decrease
2792 rolloff
2793 asr
2795 Automatic Speech Recognition
2796 This filter uses PocketSphinx for speech recognition. To enable
2798 compilation of this filter, you need to configure FFmpeg with
2799 "--enable-pocketsphinx".
2800 It accepts the following options:
2802 rate
2804 Set sampling rate of input audio. Defaults is 16000. This need to
2805 match speech models, otherwise one will get poor results.
2806 hmm Set dictionary containing acoustic model files.
2808 dict
2810 Set pronunciation dictionary.
2811 lm Set language model file.
2813 lmctl
2815 Set language model set.
2816 lmname
2818 Set which language model to use.
2819 logfn
2821 Set output for log messages.
2822 The filter exports recognized speech as the frame metadata
2824 "lavfi.asr.text".
2825 astats
2827 Display time domain statistical information about the audio channels.
2828 Statistics are calculated and displayed for each audio channel and,
2829 where applicable, an overall figure is also given.
2830 It accepts the following option:
2832 length
2834 Short window length in seconds, used for peak and trough RMS
2835 measurement. Default is 0.05 (50 milliseconds). Allowed range is
2836 "[0 - 10]".
2837 metadata
2839 Set metadata injection. All the metadata keys are prefixed with
2840 "lavfi.astats.X", where "X" is channel number starting from 1 or
2841 string "Overall". Default is disabled.
2842 Available keys for each channel are: Bit_depth Crest_factor
2844 DC_offset Dynamic_range Entropy Flat_factor Max_difference
2845 Max_level Mean_difference Min_difference Min_level Noise_floor
2846 Noise_floor_count Number_of_Infs Number_of_NaNs Number_of_denormals
2847 Peak_count Peak_level RMS_difference RMS_peak RMS_trough
2848 Zero_crossings Zero_crossings_rate
2849 and for "Overall": Bit_depth DC_offset Entropy Flat_factor
2851 Max_difference Max_level Mean_difference Min_difference Min_level
2852 Noise_floor Noise_floor_count Number_of_Infs Number_of_NaNs
2853 Number_of_denormals Number_of_samples Peak_count Peak_level
2854 RMS_difference RMS_level RMS_peak RMS_trough
2855 For example, a full key looks like "lavfi.astats.1.DC_offset" or
2857 "lavfi.astats.Overall.Peak_count".
2858 Read below for the description of the keys.
2860 reset
2862 Set the number of frames over which cumulative stats are calculated
2863 before being reset. Default is disabled.
2864 measure_perchannel
2866 Select the parameters which are measured per channel. The metadata
2867 keys can be used as flags, default is all which measures
2868 everything. none disables all per channel measurement.
2869 measure_overall
2871 Select the parameters which are measured overall. The metadata keys
2872 can be used as flags, default is all which measures everything.
2873 none disables all overall measurement.
2874 A description of the measure keys follow:
2876 none
2878 no measures
2879 all all measures
2881 Bit_depth
2883 overall bit depth of audio, i.e. number of bits used for each
2884 sample
2885 Crest_factor
2887 standard ratio of peak to RMS level (note: not in dB)
2888 DC_offset
2890 mean amplitude displacement from zero
2891 Dynamic_range
2893 measured dynamic range of audio in dB
2894 Entropy
2896 entropy measured across whole audio, entropy of value near 1.0 is
2897 typically measured for white noise
2898 Flat_factor
2900 flatness (i.e. consecutive samples with the same value) of the
2901 signal at its peak levels (i.e. either Min_level or Max_level)
2902 Max_difference
2904 maximal difference between two consecutive samples
2905 Max_level
2907 maximal sample level
2908 Mean_difference
2910 mean difference between two consecutive samples, i.e. the average
2911 of each difference between two consecutive samples
2912 Min_difference
2914 minimal difference between two consecutive samples
2915 Min_level
2917 minimal sample level
2918 Noise_floor
2920 minimum local peak measured in dBFS over a short window
2921 Noise_floor_count
2923 number of occasions (not the number of samples) that the signal
2924 attained Noise floor
2925 Number_of_Infs
2927 number of samples with an infinite value
2928 Number_of_NaNs
2930 number of samples with a NaN (not a number) value
2931 Number_of_denormals
2933 number of samples with a subnormal value
2934 Number_of_samples
2936 number of samples
2937 Peak_count
2939 number of occasions (not the number of samples) that the signal
2940 attained either Min_level or Max_level
2941 Peak_level
2943 standard peak level measured in dBFS
2944 RMS_difference
2946 Root Mean Square difference between two consecutive samples
2947 RMS_level
2949 standard RMS level measured in dBFS
2950 RMS_peak
2952 RMS_trough
2953 peak and trough values for RMS level measured over a short window,
2954 measured in dBFS.
2955 Zero crossings
2957 number of points where the waveform crosses the zero level axis
2958 Zero crossings rate
2960 rate of Zero crossings and number of audio samples
2961 asubboost
2963 Boost subwoofer frequencies.
2964 The filter accepts the following options:
2966 dry Set dry gain, how much of original signal is kept. Allowed range is
2968 from 0 to 1. Default value is 1.0.
2969 wet Set wet gain, how much of filtered signal is kept. Allowed range is
2971 from 0 to 1. Default value is 1.0.
2972 boost
2974 Set max boost factor. Allowed range is from 1 to 12. Default value
2975 is 2.
2976 decay
2978 Set delay line decay gain value. Allowed range is from 0 to 1.
2979 Default value is 0.0.
2980 feedback
2982 Set delay line feedback gain value. Allowed range is from 0 to 1.
2983 Default value is 0.9.
2984 cutoff
2986 Set cutoff frequency in Hertz. Allowed range is 50 to 900. Default
2987 value is 100.
2988 slope
2990 Set slope amount for cutoff frequency. Allowed range is 0.0001 to
2991 1. Default value is 0.5.
2992 delay
2994 Set delay. Allowed range is from 1 to 100. Default value is 20.
2995 channels
2997 Set the channels to process. Default value is all available.
2998 Commands
3000 This filter supports the all above options as commands.
3002 asubcut
3004 Cut subwoofer frequencies.
3005 This filter allows to set custom, steeper roll off than highpass
3007 filter, and thus is able to more attenuate frequency content in stop-
3008 band.
3009 The filter accepts the following options:
3011 cutoff
3013 Set cutoff frequency in Hertz. Allowed range is 2 to 200. Default
3014 value is 20.
3015 order
3017 Set filter order. Available values are from 3 to 20. Default value
3018 is 10.
3019 level
3021 Set input gain level. Allowed range is from 0 to 1. Default value
3022 is 1.
3023 Commands
3025 This filter supports the all above options as commands.
3027 asupercut
3029 Cut super frequencies.
3030 The filter accepts the following options:
3032 cutoff
3034 Set cutoff frequency in Hertz. Allowed range is 20000 to 192000.
3035 Default value is 20000.
3036 order
3038 Set filter order. Available values are from 3 to 20. Default value
3039 is 10.
3040 level
3042 Set input gain level. Allowed range is from 0 to 1. Default value
3043 is 1.
3044 Commands
3046 This filter supports the all above options as commands.
3048 asuperpass
3050 Apply high order Butterworth band-pass filter.
3051 The filter accepts the following options:
3053 centerf
3055 Set center frequency in Hertz. Allowed range is 2 to 999999.
3056 Default value is 1000.
3057 order
3059 Set filter order. Available values are from 4 to 20. Default value
3060 is 4.
3061 qfactor
3063 Set Q-factor. Allowed range is from 0.01 to 100. Default value is
3064 1.
3065 level
3067 Set input gain level. Allowed range is from 0 to 2. Default value
3068 is 1.
3069 Commands
3071 This filter supports the all above options as commands.
3073 asuperstop
3075 Apply high order Butterworth band-stop filter.
3076 The filter accepts the following options:
3078 centerf
3080 Set center frequency in Hertz. Allowed range is 2 to 999999.
3081 Default value is 1000.
3082 order
3084 Set filter order. Available values are from 4 to 20. Default value
3085 is 4.
3086 qfactor
3088 Set Q-factor. Allowed range is from 0.01 to 100. Default value is
3089 1.
3090 level
3092 Set input gain level. Allowed range is from 0 to 2. Default value
3093 is 1.
3094 Commands
3096 This filter supports the all above options as commands.
3098 atempo
3100 Adjust audio tempo.
3101 The filter accepts exactly one parameter, the audio tempo. If not
3103 specified then the filter will assume nominal 1.0 tempo. Tempo must be
3104 in the [0.5, 100.0] range.
3105 Note that tempo greater than 2 will skip some samples rather than blend
3107 them in. If for any reason this is a concern it is always possible to
3108 daisy-chain several instances of atempo to achieve the desired product
3109 tempo.
3110 Examples
3112 • Slow down audio to 80% tempo:
3114 atempo=0.8
3116 • To speed up audio to 300% tempo:
3118 atempo=3
3120 • To speed up audio to 300% tempo by daisy-chaining two atempo
3122 instances:
3123 atempo=sqrt(3),atempo=sqrt(3)
3125 Commands
3127 This filter supports the following commands:
3129 tempo
3131 Change filter tempo scale factor. Syntax for the command is :
3132 "tempo"
3133 atilt
3135 Apply spectral tilt filter to audio stream.
3136 This filter apply any spectral roll-off slope over any specified
3138 frequency band.
3139 The filter accepts the following options:
3141 freq
3143 Set central frequency of tilt in Hz. Default is 10000 Hz.
3144 slope
3146 Set slope direction of tilt. Default is 0. Allowed range is from -1
3147 to 1.
3148 width
3150 Set width of tilt. Default is 1000. Allowed range is from 100 to
3151 10000.
3152 order
3154 Set order of tilt filter.
3155 level
3157 Set input volume level. Allowed range is from 0 to 4. Defalt is 1.
3158 Commands
3160 This filter supports the all above options as commands.
3162 atrim
3164 Trim the input so that the output contains one continuous subpart of
3165 the input.
3166 It accepts the following parameters:
3168 start
3170 Timestamp (in seconds) of the start of the section to keep. I.e.
3171 the audio sample with the timestamp start will be the first sample
3172 in the output.
3173 end Specify time of the first audio sample that will be dropped, i.e.
3175 the audio sample immediately preceding the one with the timestamp
3176 end will be the last sample in the output.
3177 start_pts
3179 Same as start, except this option sets the start timestamp in
3180 samples instead of seconds.
3181 end_pts
3183 Same as end, except this option sets the end timestamp in samples
3184 instead of seconds.
3185 duration
3187 The maximum duration of the output in seconds.
3188 start_sample
3190 The number of the first sample that should be output.
3191 end_sample
3193 The number of the first sample that should be dropped.
3194 start, end, and duration are expressed as time duration specifications;
3196 see the Time duration section in the ffmpeg-utils(1) manual.
3197 Note that the first two sets of the start/end options and the duration
3199 option look at the frame timestamp, while the _sample options simply
3200 count the samples that pass through the filter. So start/end_pts and
3201 start/end_sample will give different results when the timestamps are
3202 wrong, inexact or do not start at zero. Also note that this filter does
3203 not modify the timestamps. If you wish to have the output timestamps
3204 start at zero, insert the asetpts filter after the atrim filter.
3205 If multiple start or end options are set, this filter tries to be
3207 greedy and keep all samples that match at least one of the specified
3208 constraints. To keep only the part that matches all the constraints at
3209 once, chain multiple atrim filters.
3210 The defaults are such that all the input is kept. So it is possible to
3212 set e.g. just the end values to keep everything before the specified
3213 time.
3214 Examples:
3216 • Drop everything except the second minute of input:
3218 ffmpeg -i INPUT -af atrim=60:120
3220 • Keep only the first 1000 samples:
3222 ffmpeg -i INPUT -af atrim=end_sample=1000
3224 axcorrelate
3226 Calculate normalized windowed cross-correlation between two input audio
3227 streams.
3228 Resulted samples are always between -1 and 1 inclusive. If result is 1
3230 it means two input samples are highly correlated in that selected
3231 segment. Result 0 means they are not correlated at all. If result is
3232 -1 it means two input samples are out of phase, which means they cancel
3233 each other.
3234 The filter accepts the following options:
3236 size
3238 Set size of segment over which cross-correlation is calculated.
3239 Default is 256. Allowed range is from 2 to 131072.
3240 algo
3242 Set algorithm for cross-correlation. Can be "slow" or "fast".
3243 Default is "slow". Fast algorithm assumes mean values over any
3244 given segment are always zero and thus need much less calculations
3245 to make. This is generally not true, but is valid for typical
3246 audio streams.
3247 Examples
3249 • Calculate correlation between channels in stereo audio stream:
3251 ffmpeg -i stereo.wav -af channelsplit,axcorrelate=size=1024:algo=fast correlation.wav
3253 bandpass
3255 Apply a two-pole Butterworth band-pass filter with central frequency
3256 frequency, and (3dB-point) band-width width. The csg option selects a
3257 constant skirt gain (peak gain = Q) instead of the default: constant
3258 0dB peak gain. The filter roll off at 6dB per octave (20dB per
3259 decade).
3260 The filter accepts the following options:
3262 frequency, f
3264 Set the filter's central frequency. Default is 3000.
3265 csg Constant skirt gain if set to 1. Defaults to 0.
3267 width_type, t
3269 Set method to specify band-width of filter.
3270 h Hz
3272 q Q-Factor
3274 o octave
3276 s slope
3278 k kHz
3280 width, w
3282 Specify the band-width of a filter in width_type units.
3283 mix, m
3285 How much to use filtered signal in output. Default is 1. Range is
3286 between 0 and 1.
3287 channels, c
3289 Specify which channels to filter, by default all available are
3290 filtered.
3291 normalize, n
3293 Normalize biquad coefficients, by default is disabled. Enabling it
3294 will normalize magnitude response at DC to 0dB.
3295 transform, a
3297 Set transform type of IIR filter.
3298 di
3300 dii
3301 tdi
3302 tdii
3303 latt
3304 svf
3305 zdf
3306 precision, r
3307 Set precison of filtering.
3308 auto
3310 Pick automatic sample format depending on surround filters.
3311 s16 Always use signed 16-bit.
3313 s32 Always use signed 32-bit.
3315 f32 Always use float 32-bit.
3317 f64 Always use float 64-bit.
3319 block_size, b
3321 Set block size used for reverse IIR processing. If this value is
3322 set to high enough value (higher than impulse response length
3323 truncated when reaches near zero values) filtering will become
3324 linear phase otherwise if not big enough it will just produce nasty
3325 artifacts.
3326 Note that filter delay will be exactly this many samples when set
3328 to non-zero value.
3329 Commands
3331 This filter supports the following commands:
3333 frequency, f
3335 Change bandpass frequency. Syntax for the command is : "frequency"
3336 width_type, t
3338 Change bandpass width_type. Syntax for the command is :
3339 "width_type"
3340 width, w
3342 Change bandpass width. Syntax for the command is : "width"
3343 mix, m
3345 Change bandpass mix. Syntax for the command is : "mix"
3346 bandreject
3348 Apply a two-pole Butterworth band-reject filter with central frequency
3349 frequency, and (3dB-point) band-width width. The filter roll off at
3350 6dB per octave (20dB per decade).
3351 The filter accepts the following options:
3353 frequency, f
3355 Set the filter's central frequency. Default is 3000.
3356 width_type, t
3358 Set method to specify band-width of filter.
3359 h Hz
3361 q Q-Factor
3363 o octave
3365 s slope
3367 k kHz
3369 width, w
3371 Specify the band-width of a filter in width_type units.
3372 mix, m
3374 How much to use filtered signal in output. Default is 1. Range is
3375 between 0 and 1.
3376 channels, c
3378 Specify which channels to filter, by default all available are
3379 filtered.
3380 normalize, n
3382 Normalize biquad coefficients, by default is disabled. Enabling it
3383 will normalize magnitude response at DC to 0dB.
3384 transform, a
3386 Set transform type of IIR filter.
3387 di
3389 dii
3390 tdi
3391 tdii
3392 latt
3393 svf
3394 zdf
3395 precision, r
3396 Set precison of filtering.
3397 auto
3399 Pick automatic sample format depending on surround filters.
3400 s16 Always use signed 16-bit.
3402 s32 Always use signed 32-bit.
3404 f32 Always use float 32-bit.
3406 f64 Always use float 64-bit.
3408 block_size, b
3410 Set block size used for reverse IIR processing. If this value is
3411 set to high enough value (higher than impulse response length
3412 truncated when reaches near zero values) filtering will become
3413 linear phase otherwise if not big enough it will just produce nasty
3414 artifacts.
3415 Note that filter delay will be exactly this many samples when set
3417 to non-zero value.
3418 Commands
3420 This filter supports the following commands:
3422 frequency, f
3424 Change bandreject frequency. Syntax for the command is :
3425 "frequency"
3426 width_type, t
3428 Change bandreject width_type. Syntax for the command is :
3429 "width_type"
3430 width, w
3432 Change bandreject width. Syntax for the command is : "width"
3433 mix, m
3435 Change bandreject mix. Syntax for the command is : "mix"
3436 bass, lowshelf
3438 Boost or cut the bass (lower) frequencies of the audio using a two-pole
3439 shelving filter with a response similar to that of a standard hi-fi's
3440 tone-controls. This is also known as shelving equalisation (EQ).
3441 The filter accepts the following options:
3443 gain, g
3445 Give the gain at 0 Hz. Its useful range is about -20 (for a large
3446 cut) to +20 (for a large boost). Beware of clipping when using a
3447 positive gain.
3448 frequency, f
3450 Set the filter's central frequency and so can be used to extend or
3451 reduce the frequency range to be boosted or cut. The default value
3452 is 100 Hz.
3453 width_type, t
3455 Set method to specify band-width of filter.
3456 h Hz
3458 q Q-Factor
3460 o octave
3462 s slope
3464 k kHz
3466 width, w
3468 Determine how steep is the filter's shelf transition.
3469 poles, p
3471 Set number of poles. Default is 2.
3472 mix, m
3474 How much to use filtered signal in output. Default is 1. Range is
3475 between 0 and 1.
3476 channels, c
3478 Specify which channels to filter, by default all available are
3479 filtered.
3480 normalize, n
3482 Normalize biquad coefficients, by default is disabled. Enabling it
3483 will normalize magnitude response at DC to 0dB.
3484 transform, a
3486 Set transform type of IIR filter.
3487 di
3489 dii
3490 tdi
3491 tdii
3492 latt
3493 svf
3494 zdf
3495 precision, r
3496 Set precison of filtering.
3497 auto
3499 Pick automatic sample format depending on surround filters.
3500 s16 Always use signed 16-bit.
3502 s32 Always use signed 32-bit.
3504 f32 Always use float 32-bit.
3506 f64 Always use float 64-bit.
3508 block_size, b
3510 Set block size used for reverse IIR processing. If this value is
3511 set to high enough value (higher than impulse response length
3512 truncated when reaches near zero values) filtering will become
3513 linear phase otherwise if not big enough it will just produce nasty
3514 artifacts.
3515 Note that filter delay will be exactly this many samples when set
3517 to non-zero value.
3518 Commands
3520 This filter supports the following commands:
3522 frequency, f
3524 Change bass frequency. Syntax for the command is : "frequency"
3525 width_type, t
3527 Change bass width_type. Syntax for the command is : "width_type"
3528 width, w
3530 Change bass width. Syntax for the command is : "width"
3531 gain, g
3533 Change bass gain. Syntax for the command is : "gain"
3534 mix, m
3536 Change bass mix. Syntax for the command is : "mix"
3537 biquad
3539 Apply a biquad IIR filter with the given coefficients. Where b0, b1,
3540 b2 and a0, a1, a2 are the numerator and denominator coefficients
3541 respectively. and channels, c specify which channels to filter, by
3542 default all available are filtered.
3543 Commands
3545 This filter supports the following commands:
3547 a0
3549 a1
3550 a2
3551 b0
3552 b1
3553 b2 Change biquad parameter. Syntax for the command is : "value"
3554 mix, m
3556 How much to use filtered signal in output. Default is 1. Range is
3557 between 0 and 1.
3558 channels, c
3560 Specify which channels to filter, by default all available are
3561 filtered.
3562 normalize, n
3564 Normalize biquad coefficients, by default is disabled. Enabling it
3565 will normalize magnitude response at DC to 0dB.
3566 transform, a
3568 Set transform type of IIR filter.
3569 di
3571 dii
3572 tdi
3573 tdii
3574 latt
3575 svf
3576 zdf
3577 precision, r
3578 Set precison of filtering.
3579 auto
3581 Pick automatic sample format depending on surround filters.
3582 s16 Always use signed 16-bit.
3584 s32 Always use signed 32-bit.
3586 f32 Always use float 32-bit.
3588 f64 Always use float 64-bit.
3590 block_size, b
3592 Set block size used for reverse IIR processing. If this value is
3593 set to high enough value (higher than impulse response length
3594 truncated when reaches near zero values) filtering will become
3595 linear phase otherwise if not big enough it will just produce nasty
3596 artifacts.
3597 Note that filter delay will be exactly this many samples when set
3599 to non-zero value.
3600 bs2b
3602 Bauer stereo to binaural transformation, which improves headphone
3603 listening of stereo audio records.
3604 To enable compilation of this filter you need to configure FFmpeg with
3606 "--enable-libbs2b".
3607 It accepts the following parameters:
3609 profile
3611 Pre-defined crossfeed level.
3612 default
3614 Default level (fcut=700, feed=50).
3615 cmoy
3617 Chu Moy circuit (fcut=700, feed=60).
3618 jmeier
3620 Jan Meier circuit (fcut=650, feed=95).
3621 fcut
3623 Cut frequency (in Hz).
3624 feed
3626 Feed level (in Hz).
3627 channelmap
3629 Remap input channels to new locations.
3630 It accepts the following parameters:
3632 map Map channels from input to output. The argument is a '|'-separated
3634 list of mappings, each in the "in_channel-out_channel" or
3635 in_channel form. in_channel can be either the name of the input
3636 channel (e.g. FL for front left) or its index in the input channel
3637 layout. out_channel is the name of the output channel or its index
3638 in the output channel layout. If out_channel is not given then it
3639 is implicitly an index, starting with zero and increasing by one
3640 for each mapping.
3641 channel_layout
3643 The channel layout of the output stream.
3644 If no mapping is present, the filter will implicitly map input channels
3646 to output channels, preserving indices.
3647 Examples
3649 • For example, assuming a 5.1+downmix input MOV file,
3651 ffmpeg -i in.mov -filter 'channelmap=map=DL-FL|DR-FR' out.wav
3653 will create an output WAV file tagged as stereo from the downmix
3655 channels of the input.
3656 • To fix a 5.1 WAV improperly encoded in AAC's native channel order
3658 ffmpeg -i in.wav -filter 'channelmap=1|2|0|5|3|4:5.1' out.wav
3660 channelsplit
3662 Split each channel from an input audio stream into a separate output
3663 stream.
3664 It accepts the following parameters:
3666 channel_layout
3668 The channel layout of the input stream. The default is "stereo".
3669 channels
3671 A channel layout describing the channels to be extracted as
3672 separate output streams or "all" to extract each input channel as a
3673 separate stream. The default is "all".
3674 Choosing channels not present in channel layout in the input will
3676 result in an error.
3677 Examples
3679 • For example, assuming a stereo input MP3 file,
3681 ffmpeg -i in.mp3 -filter_complex channelsplit out.mkv
3683 will create an output Matroska file with two audio streams, one
3685 containing only the left channel and the other the right channel.
3686 • Split a 5.1 WAV file into per-channel files:
3688 ffmpeg -i in.wav -filter_complex
3690 'channelsplit=channel_layout=5.1[FL][FR][FC][LFE][SL][SR]'
3691 -map '[FL]' front_left.wav -map '[FR]' front_right.wav -map '[FC]'
3692 front_center.wav -map '[LFE]' lfe.wav -map '[SL]' side_left.wav -map '[SR]'
3693 side_right.wav
3694 • Extract only LFE from a 5.1 WAV file:
3696 ffmpeg -i in.wav -filter_complex 'channelsplit=channel_layout=5.1:channels=LFE[LFE]'
3698 -map '[LFE]' lfe.wav
3699 chorus
3701 Add a chorus effect to the audio.
3702 Can make a single vocal sound like a chorus, but can also be applied to
3704 instrumentation.
3705 Chorus resembles an echo effect with a short delay, but whereas with
3707 echo the delay is constant, with chorus, it is varied using using
3708 sinusoidal or triangular modulation. The modulation depth defines the
3709 range the modulated delay is played before or after the delay. Hence
3710 the delayed sound will sound slower or faster, that is the delayed
3711 sound tuned around the original one, like in a chorus where some vocals
3712 are slightly off key.
3713 It accepts the following parameters:
3715 in_gain
3717 Set input gain. Default is 0.4.
3718 out_gain
3720 Set output gain. Default is 0.4.
3721 delays
3723 Set delays. A typical delay is around 40ms to 60ms.
3724 decays
3726 Set decays.
3727 speeds
3729 Set speeds.
3730 depths
3732 Set depths.
3733 Examples
3735 • A single delay:
3737 chorus=0.7:0.9:55:0.4:0.25:2
3739 • Two delays:
3741 chorus=0.6:0.9:50|60:0.4|0.32:0.25|0.4:2|1.3
3743 • Fuller sounding chorus with three delays:
3745 chorus=0.5:0.9:50|60|40:0.4|0.32|0.3:0.25|0.4|0.3:2|2.3|1.3
3747 compand
3749 Compress or expand the audio's dynamic range.
3750 It accepts the following parameters:
3752 attacks
3754 decays
3755 A list of times in seconds for each channel over which the
3756 instantaneous level of the input signal is averaged to determine
3757 its volume. attacks refers to increase of volume and decays refers
3758 to decrease of volume. For most situations, the attack time
3759 (response to the audio getting louder) should be shorter than the
3760 decay time, because the human ear is more sensitive to sudden loud
3761 audio than sudden soft audio. A typical value for attack is 0.3
3762 seconds and a typical value for decay is 0.8 seconds. If specified
3763 number of attacks & decays is lower than number of channels, the
3764 last set attack/decay will be used for all remaining channels.
3765 points
3767 A list of points for the transfer function, specified in dB
3768 relative to the maximum possible signal amplitude. Each key points
3769 list must be defined using the following syntax:
3770 "x0/y0|x1/y1|x2/y2|...." or "x0/y0 x1/y1 x2/y2 ...."
3771 The input values must be in strictly increasing order but the
3773 transfer function does not have to be monotonically rising. The
3774 point "0/0" is assumed but may be overridden (by "0/out-dBn").
3775 Typical values for the transfer function are "-70/-70|-60/-20|1/0".
3776 soft-knee
3778 Set the curve radius in dB for all joints. It defaults to 0.01.
3779 gain
3781 Set the additional gain in dB to be applied at all points on the
3782 transfer function. This allows for easy adjustment of the overall
3783 gain. It defaults to 0.
3784 volume
3786 Set an initial volume, in dB, to be assumed for each channel when
3787 filtering starts. This permits the user to supply a nominal level
3788 initially, so that, for example, a very large gain is not applied
3789 to initial signal levels before the companding has begun to
3790 operate. A typical value for audio which is initially quiet is -90
3791 dB. It defaults to 0.
3792 delay
3794 Set a delay, in seconds. The input audio is analyzed immediately,
3795 but audio is delayed before being fed to the volume adjuster.
3796 Specifying a delay approximately equal to the attack/decay times
3797 allows the filter to effectively operate in predictive rather than
3798 reactive mode. It defaults to 0.
3799 Examples
3801 • Make music with both quiet and loud passages suitable for listening
3803 to in a noisy environment:
3804 compand=.3|.3:1|1:-90/-60|-60/-40|-40/-30|-20/-20:6:0:-90:0.2
3806 Another example for audio with whisper and explosion parts:
3808 compand=0|0:1|1:-90/-900|-70/-70|-30/-9|0/-3:6:0:0:0
3810 • A noise gate for when the noise is at a lower level than the
3812 signal:
3813 compand=.1|.1:.2|.2:-900/-900|-50.1/-900|-50/-50:.01:0:-90:.1
3815 • Here is another noise gate, this time for when the noise is at a
3817 higher level than the signal (making it, in some ways, similar to
3818 squelch):
3819 compand=.1|.1:.1|.1:-45.1/-45.1|-45/-900|0/-900:.01:45:-90:.1
3821 • 2:1 compression starting at -6dB:
3823 compand=points=-80/-80|-6/-6|0/-3.8|20/3.5
3825 • 2:1 compression starting at -9dB:
3827 compand=points=-80/-80|-9/-9|0/-5.3|20/2.9
3829 • 2:1 compression starting at -12dB:
3831 compand=points=-80/-80|-12/-12|0/-6.8|20/1.9
3833 • 2:1 compression starting at -18dB:
3835 compand=points=-80/-80|-18/-18|0/-9.8|20/0.7
3837 • 3:1 compression starting at -15dB:
3839 compand=points=-80/-80|-15/-15|0/-10.8|20/-5.2
3841 • Compressor/Gate:
3843 compand=points=-80/-105|-62/-80|-15.4/-15.4|0/-12|20/-7.6
3845 • Expander:
3847 compand=attacks=0:points=-80/-169|-54/-80|-49.5/-64.6|-41.1/-41.1|-25.8/-15|-10.8/-4.5|0/0|20/8.3
3849 • Hard limiter at -6dB:
3851 compand=attacks=0:points=-80/-80|-6/-6|20/-6
3853 • Hard limiter at -12dB:
3855 compand=attacks=0:points=-80/-80|-12/-12|20/-12
3857 • Hard noise gate at -35 dB:
3859 compand=attacks=0:points=-80/-115|-35.1/-80|-35/-35|20/20
3861 • Soft limiter:
3863 compand=attacks=0:points=-80/-80|-12.4/-12.4|-6/-8|0/-6.8|20/-2.8
3865 compensationdelay
3867 Compensation Delay Line is a metric based delay to compensate differing
3868 positions of microphones or speakers.
3869 For example, you have recorded guitar with two microphones placed in
3871 different locations. Because the front of sound wave has fixed speed in
3872 normal conditions, the phasing of microphones can vary and depends on
3873 their location and interposition. The best sound mix can be achieved
3874 when these microphones are in phase (synchronized). Note that a
3875 distance of ~30 cm between microphones makes one microphone capture the
3876 signal in antiphase to the other microphone. That makes the final mix
3877 sound moody. This filter helps to solve phasing problems by adding
3878 different delays to each microphone track and make them synchronized.
3879 The best result can be reached when you take one track as base and
3881 synchronize other tracks one by one with it. Remember that
3882 synchronization/delay tolerance depends on sample rate, too. Higher
3883 sample rates will give more tolerance.
3884 The filter accepts the following parameters:
3886 mm Set millimeters distance. This is compensation distance for fine
3888 tuning. Default is 0.
3889 cm Set cm distance. This is compensation distance for tightening
3891 distance setup. Default is 0.
3892 m Set meters distance. This is compensation distance for hard
3894 distance setup. Default is 0.
3895 dry Set dry amount. Amount of unprocessed (dry) signal. Default is 0.
3897 wet Set wet amount. Amount of processed (wet) signal. Default is 1.
3899 temp
3901 Set temperature in degrees Celsius. This is the temperature of the
3902 environment. Default is 20.
3903 Commands
3905 This filter supports the all above options as commands.
3907 crossfeed
3909 Apply headphone crossfeed filter.
3910 Crossfeed is the process of blending the left and right channels of
3912 stereo audio recording. It is mainly used to reduce extreme stereo
3913 separation of low frequencies.
3914 The intent is to produce more speaker like sound to the listener.
3916 The filter accepts the following options:
3918 strength
3920 Set strength of crossfeed. Default is 0.2. Allowed range is from 0
3921 to 1. This sets gain of low shelf filter for side part of stereo
3922 image. Default is -6dB. Max allowed is -30db when strength is set
3923 to 1.
3924 range
3926 Set soundstage wideness. Default is 0.5. Allowed range is from 0 to
3927 1. This sets cut off frequency of low shelf filter. Default is cut
3928 off near 1550 Hz. With range set to 1 cut off frequency is set to
3929 2100 Hz.
3930 slope
3932 Set curve slope of low shelf filter. Default is 0.5. Allowed range
3933 is from 0.01 to 1.
3934 level_in
3936 Set input gain. Default is 0.9.
3937 level_out
3939 Set output gain. Default is 1.
3940 block_size
3942 Set block size used for reverse IIR processing. If this value is
3943 set to high enough value (higher than impulse response length
3944 truncated when reaches near zero values) filtering will become
3945 linear phase otherwise if not big enough it will just produce nasty
3946 artifacts.
3947 Note that filter delay will be exactly this many samples when set
3949 to non-zero value.
3950 Commands
3952 This filter supports the all above options as commands.
3954 crystalizer
3956 Simple algorithm for audio noise sharpening.
3957 This filter linearly increases differences betweeen each audio sample.
3959 The filter accepts the following options:
3961 i Sets the intensity of effect (default: 2.0). Must be in range
3963 between -10.0 to 0 (unchanged sound) to 10.0 (maximum effect). To
3964 inverse filtering use negative value.
3965 c Enable clipping. By default is enabled.
3967 Commands
3969 This filter supports the all above options as commands.
3971 dcshift
3973 Apply a DC shift to the audio.
3974 This can be useful to remove a DC offset (caused perhaps by a hardware
3976 problem in the recording chain) from the audio. The effect of a DC
3977 offset is reduced headroom and hence volume. The astats filter can be
3978 used to determine if a signal has a DC offset.
3979 shift
3981 Set the DC shift, allowed range is [-1, 1]. It indicates the amount
3982 to shift the audio.
3983 limitergain
3985 Optional. It should have a value much less than 1 (e.g. 0.05 or
3986 0.02) and is used to prevent clipping.
3987 deesser
3989 Apply de-essing to the audio samples.
3990 i Set intensity for triggering de-essing. Allowed range is from 0 to
3992 1. Default is 0.
3993 m Set amount of ducking on treble part of sound. Allowed range is
3995 from 0 to 1. Default is 0.5.
3996 f How much of original frequency content to keep when de-essing.
3998 Allowed range is from 0 to 1. Default is 0.5.
3999 s Set the output mode.
4001 It accepts the following values:
4003 i Pass input unchanged.
4005 o Pass ess filtered out.
4007 e Pass only ess.
4009 Default value is o.
4011 dialoguenhance
4013 Enhance dialogue in stereo audio.
4014 This filter accepts stereo input and produce surround (3.0) channels
4016 output. The newly produced front center channel have enhanced speech
4017 dialogue originally available in both stereo channels. This filter
4018 outputs front left and front right channels same as available in stereo
4019 input.
4020 The filter accepts the following options:
4022 original
4024 Set the original center factor to keep in front center channel
4025 output. Allowed range is from 0 to 1. Default value is 1.
4026 enhance
4028 Set the dialogue enhance factor to put in front center channel
4029 output. Allowed range is from 0 to 3. Default value is 1.
4030 voice
4032 Set the voice detection factor. Allowed range is from 2 to 32.
4033 Default value is 2.
4034 Commands
4036 This filter supports the all above options as commands.
4038 drmeter
4040 Measure audio dynamic range.
4041 DR values of 14 and higher is found in very dynamic material. DR of 8
4043 to 13 is found in transition material. And anything less that 8 have
4044 very poor dynamics and is very compressed.
4045 The filter accepts the following options:
4047 length
4049 Set window length in seconds used to split audio into segments of
4050 equal length. Default is 3 seconds.
4051 dynaudnorm
4053 Dynamic Audio Normalizer.
4054 This filter applies a certain amount of gain to the input audio in
4056 order to bring its peak magnitude to a target level (e.g. 0 dBFS).
4057 However, in contrast to more "simple" normalization algorithms, the
4058 Dynamic Audio Normalizer *dynamically* re-adjusts the gain factor to
4059 the input audio. This allows for applying extra gain to the "quiet"
4060 sections of the audio while avoiding distortions or clipping the "loud"
4061 sections. In other words: The Dynamic Audio Normalizer will "even out"
4062 the volume of quiet and loud sections, in the sense that the volume of
4063 each section is brought to the same target level. Note, however, that
4064 the Dynamic Audio Normalizer achieves this goal *without* applying
4065 "dynamic range compressing". It will retain 100% of the dynamic range
4066 *within* each section of the audio file.
4067 framelen, f
4069 Set the frame length in milliseconds. In range from 10 to 8000
4070 milliseconds. Default is 500 milliseconds. The Dynamic Audio
4071 Normalizer processes the input audio in small chunks, referred to
4072 as frames. This is required, because a peak magnitude has no
4073 meaning for just a single sample value. Instead, we need to
4074 determine the peak magnitude for a contiguous sequence of sample
4075 values. While a "standard" normalizer would simply use the peak
4076 magnitude of the complete file, the Dynamic Audio Normalizer
4077 determines the peak magnitude individually for each frame. The
4078 length of a frame is specified in milliseconds. By default, the
4079 Dynamic Audio Normalizer uses a frame length of 500 milliseconds,
4080 which has been found to give good results with most files. Note
4081 that the exact frame length, in number of samples, will be
4082 determined automatically, based on the sampling rate of the
4083 individual input audio file.
4084 gausssize, g
4086 Set the Gaussian filter window size. In range from 3 to 301, must
4087 be odd number. Default is 31. Probably the most important
4088 parameter of the Dynamic Audio Normalizer is the "window size" of
4089 the Gaussian smoothing filter. The filter's window size is
4090 specified in frames, centered around the current frame. For the
4091 sake of simplicity, this must be an odd number. Consequently, the
4092 default value of 31 takes into account the current frame, as well
4093 as the 15 preceding frames and the 15 subsequent frames. Using a
4094 larger window results in a stronger smoothing effect and thus in
4095 less gain variation, i.e. slower gain adaptation. Conversely, using
4096 a smaller window results in a weaker smoothing effect and thus in
4097 more gain variation, i.e. faster gain adaptation. In other words,
4098 the more you increase this value, the more the Dynamic Audio
4099 Normalizer will behave like a "traditional" normalization filter.
4100 On the contrary, the more you decrease this value, the more the
4101 Dynamic Audio Normalizer will behave like a dynamic range
4102 compressor.
4103 peak, p
4105 Set the target peak value. This specifies the highest permissible
4106 magnitude level for the normalized audio input. This filter will
4107 try to approach the target peak magnitude as closely as possible,
4108 but at the same time it also makes sure that the normalized signal
4109 will never exceed the peak magnitude. A frame's maximum local gain
4110 factor is imposed directly by the target peak magnitude. The
4111 default value is 0.95 and thus leaves a headroom of 5%*. It is not
4112 recommended to go above this value.
4113 maxgain, m
4115 Set the maximum gain factor. In range from 1.0 to 100.0. Default is
4116 10.0. The Dynamic Audio Normalizer determines the maximum possible
4117 (local) gain factor for each input frame, i.e. the maximum gain
4118 factor that does not result in clipping or distortion. The maximum
4119 gain factor is determined by the frame's highest magnitude sample.
4120 However, the Dynamic Audio Normalizer additionally bounds the
4121 frame's maximum gain factor by a predetermined (global) maximum
4122 gain factor. This is done in order to avoid excessive gain factors
4123 in "silent" or almost silent frames. By default, the maximum gain
4124 factor is 10.0, For most inputs the default value should be
4125 sufficient and it usually is not recommended to increase this
4126 value. Though, for input with an extremely low overall volume
4127 level, it may be necessary to allow even higher gain factors. Note,
4128 however, that the Dynamic Audio Normalizer does not simply apply a
4129 "hard" threshold (i.e. cut off values above the threshold).
4130 Instead, a "sigmoid" threshold function will be applied. This way,
4131 the gain factors will smoothly approach the threshold value, but
4132 never exceed that value.
4133 targetrms, r
4135 Set the target RMS. In range from 0.0 to 1.0. Default is 0.0 -
4136 disabled. By default, the Dynamic Audio Normalizer performs "peak"
4137 normalization. This means that the maximum local gain factor for
4138 each frame is defined (only) by the frame's highest magnitude
4139 sample. This way, the samples can be amplified as much as possible
4140 without exceeding the maximum signal level, i.e. without clipping.
4141 Optionally, however, the Dynamic Audio Normalizer can also take
4142 into account the frame's root mean square, abbreviated RMS. In
4143 electrical engineering, the RMS is commonly used to determine the
4144 power of a time-varying signal. It is therefore considered that the
4145 RMS is a better approximation of the "perceived loudness" than just
4146 looking at the signal's peak magnitude. Consequently, by adjusting
4147 all frames to a constant RMS value, a uniform "perceived loudness"
4148 can be established. If a target RMS value has been specified, a
4149 frame's local gain factor is defined as the factor that would
4150 result in exactly that RMS value. Note, however, that the maximum
4151 local gain factor is still restricted by the frame's highest
4152 magnitude sample, in order to prevent clipping.
4153 coupling, n
4155 Enable channels coupling. By default is enabled. By default, the
4156 Dynamic Audio Normalizer will amplify all channels by the same
4157 amount. This means the same gain factor will be applied to all
4158 channels, i.e. the maximum possible gain factor is determined by
4159 the "loudest" channel. However, in some recordings, it may happen
4160 that the volume of the different channels is uneven, e.g. one
4161 channel may be "quieter" than the other one(s). In this case, this
4162 option can be used to disable the channel coupling. This way, the
4163 gain factor will be determined independently for each channel,
4164 depending only on the individual channel's highest magnitude
4165 sample. This allows for harmonizing the volume of the different
4166 channels.
4167 correctdc, c
4169 Enable DC bias correction. By default is disabled. An audio signal
4170 (in the time domain) is a sequence of sample values. In the
4171 Dynamic Audio Normalizer these sample values are represented in the
4172 -1.0 to 1.0 range, regardless of the original input format.
4173 Normally, the audio signal, or "waveform", should be centered
4174 around the zero point. That means if we calculate the mean value
4175 of all samples in a file, or in a single frame, then the result
4176 should be 0.0 or at least very close to that value. If, however,
4177 there is a significant deviation of the mean value from 0.0, in
4178 either positive or negative direction, this is referred to as a DC
4179 bias or DC offset. Since a DC bias is clearly undesirable, the
4180 Dynamic Audio Normalizer provides optional DC bias correction.
4181 With DC bias correction enabled, the Dynamic Audio Normalizer will
4182 determine the mean value, or "DC correction" offset, of each input
4183 frame and subtract that value from all of the frame's sample values
4184 which ensures those samples are centered around 0.0 again. Also, in
4185 order to avoid "gaps" at the frame boundaries, the DC correction
4186 offset values will be interpolated smoothly between neighbouring
4187 frames.
4188 altboundary, b
4190 Enable alternative boundary mode. By default is disabled. The
4191 Dynamic Audio Normalizer takes into account a certain neighbourhood
4192 around each frame. This includes the preceding frames as well as
4193 the subsequent frames. However, for the "boundary" frames, located
4194 at the very beginning and at the very end of the audio file, not
4195 all neighbouring frames are available. In particular, for the first
4196 few frames in the audio file, the preceding frames are not known.
4197 And, similarly, for the last few frames in the audio file, the
4198 subsequent frames are not known. Thus, the question arises which
4199 gain factors should be assumed for the missing frames in the
4200 "boundary" region. The Dynamic Audio Normalizer implements two
4201 modes to deal with this situation. The default boundary mode
4202 assumes a gain factor of exactly 1.0 for the missing frames,
4203 resulting in a smooth "fade in" and "fade out" at the beginning and
4204 at the end of the input, respectively.
4205 compress, s
4207 Set the compress factor. In range from 0.0 to 30.0. Default is 0.0.
4208 By default, the Dynamic Audio Normalizer does not apply
4209 "traditional" compression. This means that signal peaks will not be
4210 pruned and thus the full dynamic range will be retained within each
4211 local neighbourhood. However, in some cases it may be desirable to
4212 combine the Dynamic Audio Normalizer's normalization algorithm with
4213 a more "traditional" compression. For this purpose, the Dynamic
4214 Audio Normalizer provides an optional compression (thresholding)
4215 function. If (and only if) the compression feature is enabled, all
4216 input frames will be processed by a soft knee thresholding function
4217 prior to the actual normalization process. Put simply, the
4218 thresholding function is going to prune all samples whose magnitude
4219 exceeds a certain threshold value. However, the Dynamic Audio
4220 Normalizer does not simply apply a fixed threshold value. Instead,
4221 the threshold value will be adjusted for each individual frame. In
4222 general, smaller parameters result in stronger compression, and
4223 vice versa. Values below 3.0 are not recommended, because audible
4224 distortion may appear.
4225 threshold, t
4227 Set the target threshold value. This specifies the lowest
4228 permissible magnitude level for the audio input which will be
4229 normalized. If input frame volume is above this value frame will
4230 be normalized. Otherwise frame may not be normalized at all. The
4231 default value is set to 0, which means all input frames will be
4232 normalized. This option is mostly useful if digital noise is not
4233 wanted to be amplified.
4234 channels, h
4236 Specify which channels to filter, by default all available channels
4237 are filtered.
4238 overlap, o
4240 Specify overlap for frames. If set to 0 (default) no frame
4241 overlapping is done. Using >0 and <1 values will make less
4242 conservative gain adjustments, like when framelen option is set to
4243 smaller value, if framelen option value is compensated for non-zero
4244 overlap then gain adjustments will be smoother across time compared
4245 to zero overlap case.
4246 curve, v
4248 Specify the peak mapping curve expression which is going to be used
4249 when calculating gain applied to frames. The max output frame gain
4250 will still be limited by other options mentioned previously for
4251 this filter.
4252 The expression can contain the following constants:
4254 ch current channel number
4256 sn current sample number
4258 nb_channels
4260 number of channels
4261 t timestamp expressed in seconds
4263 sr sample rate
4265 p current frame peak value
4267 Commands
4269 This filter supports the all above options as commands.
4271 earwax
4273 Make audio easier to listen to on headphones.
4274 This filter adds `cues' to 44.1kHz stereo (i.e. audio CD format) audio
4276 so that when listened to on headphones the stereo image is moved from
4277 inside your head (standard for headphones) to outside and in front of
4278 the listener (standard for speakers).
4279 Ported from SoX.
4281 equalizer
4283 Apply a two-pole peaking equalisation (EQ) filter. With this filter,
4284 the signal-level at and around a selected frequency can be increased or
4285 decreased, whilst (unlike bandpass and bandreject filters) that at all
4286 other frequencies is unchanged.
4287 In order to produce complex equalisation curves, this filter can be
4289 given several times, each with a different central frequency.
4290 The filter accepts the following options:
4292 frequency, f
4294 Set the filter's central frequency in Hz.
4295 width_type, t
4297 Set method to specify band-width of filter.
4298 h Hz
4300 q Q-Factor
4302 o octave
4304 s slope
4306 k kHz
4308 width, w
4310 Specify the band-width of a filter in width_type units.
4311 gain, g
4313 Set the required gain or attenuation in dB. Beware of clipping
4314 when using a positive gain.
4315 mix, m
4317 How much to use filtered signal in output. Default is 1. Range is
4318 between 0 and 1.
4319 channels, c
4321 Specify which channels to filter, by default all available are
4322 filtered.
4323 normalize, n
4325 Normalize biquad coefficients, by default is disabled. Enabling it
4326 will normalize magnitude response at DC to 0dB.
4327 transform, a
4329 Set transform type of IIR filter.
4330 di
4332 dii
4333 tdi
4334 tdii
4335 latt
4336 svf
4337 zdf
4338 precision, r
4339 Set precison of filtering.
4340 auto
4342 Pick automatic sample format depending on surround filters.
4343 s16 Always use signed 16-bit.
4345 s32 Always use signed 32-bit.
4347 f32 Always use float 32-bit.
4349 f64 Always use float 64-bit.
4351 block_size, b
4353 Set block size used for reverse IIR processing. If this value is
4354 set to high enough value (higher than impulse response length
4355 truncated when reaches near zero values) filtering will become
4356 linear phase otherwise if not big enough it will just produce nasty
4357 artifacts.
4358 Note that filter delay will be exactly this many samples when set
4360 to non-zero value.
4361 Examples
4363 • Attenuate 10 dB at 1000 Hz, with a bandwidth of 200 Hz:
4365 equalizer=f=1000:t=h:width=200:g=-10
4367 • Apply 2 dB gain at 1000 Hz with Q 1 and attenuate 5 dB at 100 Hz
4369 with Q 2:
4370 equalizer=f=1000:t=q:w=1:g=2,equalizer=f=100:t=q:w=2:g=-5
4372 Commands
4374 This filter supports the following commands:
4376 frequency, f
4378 Change equalizer frequency. Syntax for the command is :
4379 "frequency"
4380 width_type, t
4382 Change equalizer width_type. Syntax for the command is :
4383 "width_type"
4384 width, w
4386 Change equalizer width. Syntax for the command is : "width"
4387 gain, g
4389 Change equalizer gain. Syntax for the command is : "gain"
4390 mix, m
4392 Change equalizer mix. Syntax for the command is : "mix"
4393 extrastereo
4395 Linearly increases the difference between left and right channels which
4396 adds some sort of "live" effect to playback.
4397 The filter accepts the following options:
4399 m Sets the difference coefficient (default: 2.5). 0.0 means mono
4401 sound (average of both channels), with 1.0 sound will be unchanged,
4402 with -1.0 left and right channels will be swapped.
4403 c Enable clipping. By default is enabled.
4405 Commands
4407 This filter supports the all above options as commands.
4409 firequalizer
4411 Apply FIR Equalization using arbitrary frequency response.
4412 The filter accepts the following option:
4414 gain
4416 Set gain curve equation (in dB). The expression can contain
4417 variables:
4418 f the evaluated frequency
4420 sr sample rate
4422 ch channel number, set to 0 when multichannels evaluation is
4424 disabled
4425 chid
4427 channel id, see libavutil/channel_layout.h, set to the first
4428 channel id when multichannels evaluation is disabled
4429 chs number of channels
4431 chlayout
4433 channel_layout, see libavutil/channel_layout.h
4434 and functions:
4436 gain_interpolate(f)
4438 interpolate gain on frequency f based on gain_entry
4439 cubic_interpolate(f)
4441 same as gain_interpolate, but smoother
4442 This option is also available as command. Default is
4444 gain_interpolate(f).
4445 gain_entry
4447 Set gain entry for gain_interpolate function. The expression can
4448 contain functions:
4449 entry(f, g)
4451 store gain entry at frequency f with value g
4452 This option is also available as command.
4454 delay
4456 Set filter delay in seconds. Higher value means more accurate.
4457 Default is 0.01.
4458 accuracy
4460 Set filter accuracy in Hz. Lower value means more accurate.
4461 Default is 5.
4462 wfunc
4464 Set window function. Acceptable values are:
4465 rectangular
4467 rectangular window, useful when gain curve is already smooth
4468 hann
4470 hann window (default)
4471 hamming
4473 hamming window
4474 blackman
4476 blackman window
4477 nuttall3
4479 3-terms continuous 1st derivative nuttall window
4480 mnuttall3
4482 minimum 3-terms discontinuous nuttall window
4483 nuttall
4485 4-terms continuous 1st derivative nuttall window
4486 bnuttall
4488 minimum 4-terms discontinuous nuttall (blackman-nuttall) window
4489 bharris
4491 blackman-harris window
4492 tukey
4494 tukey window
4495 fixed
4497 If enabled, use fixed number of audio samples. This improves speed
4498 when filtering with large delay. Default is disabled.
4499 multi
4501 Enable multichannels evaluation on gain. Default is disabled.
4502 zero_phase
4504 Enable zero phase mode by subtracting timestamp to compensate
4505 delay. Default is disabled.
4506 scale
4508 Set scale used by gain. Acceptable values are:
4509 linlin
4511 linear frequency, linear gain
4512 linlog
4514 linear frequency, logarithmic (in dB) gain (default)
4515 loglin
4517 logarithmic (in octave scale where 20 Hz is 0) frequency,
4518 linear gain
4519 loglog
4521 logarithmic frequency, logarithmic gain
4522 dumpfile
4524 Set file for dumping, suitable for gnuplot.
4525 dumpscale
4527 Set scale for dumpfile. Acceptable values are same with scale
4528 option. Default is linlog.
4529 fft2
4531 Enable 2-channel convolution using complex FFT. This improves speed
4532 significantly. Default is disabled.
4533 min_phase
4535 Enable minimum phase impulse response. Default is disabled.
4536 Examples
4538 • lowpass at 1000 Hz:
4540 firequalizer=gain='if(lt(f,1000), 0, -INF)'
4542 • lowpass at 1000 Hz with gain_entry:
4544 firequalizer=gain_entry='entry(1000,0); entry(1001, -INF)'
4546 • custom equalization:
4548 firequalizer=gain_entry='entry(100,0); entry(400, -4); entry(1000, -6); entry(2000, 0)'
4550 • higher delay with zero phase to compensate delay:
4552 firequalizer=delay=0.1:fixed=on:zero_phase=on
4554 • lowpass on left channel, highpass on right channel:
4556 firequalizer=gain='if(eq(chid,1), gain_interpolate(f), if(eq(chid,2), gain_interpolate(1e6+f), 0))'
4558 :gain_entry='entry(1000, 0); entry(1001,-INF); entry(1e6+1000,0)':multi=on
4559 flanger
4561 Apply a flanging effect to the audio.
4562 The filter accepts the following options:
4564 delay
4566 Set base delay in milliseconds. Range from 0 to 30. Default value
4567 is 0.
4568 depth
4570 Set added sweep delay in milliseconds. Range from 0 to 10. Default
4571 value is 2.
4572 regen
4574 Set percentage regeneration (delayed signal feedback). Range from
4575 -95 to 95. Default value is 0.
4576 width
4578 Set percentage of delayed signal mixed with original. Range from 0
4579 to 100. Default value is 71.
4580 speed
4582 Set sweeps per second (Hz). Range from 0.1 to 10. Default value is
4583 0.5.
4584 shape
4586 Set swept wave shape, can be triangular or sinusoidal. Default
4587 value is sinusoidal.
4588 phase
4590 Set swept wave percentage-shift for multi channel. Range from 0 to
4591 100. Default value is 25.
4592 interp
4594 Set delay-line interpolation, linear or quadratic. Default is
4595 linear.
4596 haas
4598 Apply Haas effect to audio.
4599 Note that this makes most sense to apply on mono signals. With this
4601 filter applied to mono signals it give some directionality and
4602 stretches its stereo image.
4603 The filter accepts the following options:
4605 level_in
4607 Set input level. By default is 1, or 0dB
4608 level_out
4610 Set output level. By default is 1, or 0dB.
4611 side_gain
4613 Set gain applied to side part of signal. By default is 1.
4614 middle_source
4616 Set kind of middle source. Can be one of the following:
4617 left
4619 Pick left channel.
4620 right
4622 Pick right channel.
4623 mid Pick middle part signal of stereo image.
4625 side
4627 Pick side part signal of stereo image.
4628 middle_phase
4630 Change middle phase. By default is disabled.
4631 left_delay
4633 Set left channel delay. By default is 2.05 milliseconds.
4634 left_balance
4636 Set left channel balance. By default is -1.
4637 left_gain
4639 Set left channel gain. By default is 1.
4640 left_phase
4642 Change left phase. By default is disabled.
4643 right_delay
4645 Set right channel delay. By defaults is 2.12 milliseconds.
4646 right_balance
4648 Set right channel balance. By default is 1.
4649 right_gain
4651 Set right channel gain. By default is 1.
4652 right_phase
4654 Change right phase. By default is enabled.
4655 hdcd
4657 Decodes High Definition Compatible Digital (HDCD) data. A 16-bit PCM
4658 stream with embedded HDCD codes is expanded into a 20-bit PCM stream.
4659 The filter supports the Peak Extend and Low-level Gain Adjustment
4661 features of HDCD, and detects the Transient Filter flag.
4662 ffmpeg -i HDCD16.flac -af hdcd OUT24.flac
4664 When using the filter with wav, note the default encoding for wav is
4666 16-bit, so the resulting 20-bit stream will be truncated back to
4667 16-bit. Use something like -acodec pcm_s24le after the filter to get
4668 24-bit PCM output.
4669 ffmpeg -i HDCD16.wav -af hdcd OUT16.wav
4671 ffmpeg -i HDCD16.wav -af hdcd -c:a pcm_s24le OUT24.wav
4672 The filter accepts the following options:
4674 disable_autoconvert
4676 Disable any automatic format conversion or resampling in the filter
4677 graph.
4678 process_stereo
4680 Process the stereo channels together. If target_gain does not match
4681 between channels, consider it invalid and use the last valid
4682 target_gain.
4683 cdt_ms
4685 Set the code detect timer period in ms.
4686 force_pe
4688 Always extend peaks above -3dBFS even if PE isn't signaled.
4689 analyze_mode
4691 Replace audio with a solid tone and adjust the amplitude to signal
4692 some specific aspect of the decoding process. The output file can
4693 be loaded in an audio editor alongside the original to aid
4694 analysis.
4695 "analyze_mode=pe:force_pe=true" can be used to see all samples
4697 above the PE level.
4698 Modes are:
4700 0, off
4702 Disabled
4703 1, lle
4705 Gain adjustment level at each sample
4706 2, pe
4708 Samples where peak extend occurs
4709 3, cdt
4711 Samples where the code detect timer is active
4712 4, tgm
4714 Samples where the target gain does not match between channels
4715 headphone
4717 Apply head-related transfer functions (HRTFs) to create virtual
4718 loudspeakers around the user for binaural listening via headphones.
4719 The HRIRs are provided via additional streams, for each channel one
4720 stereo input stream is needed.
4721 The filter accepts the following options:
4723 map Set mapping of input streams for convolution. The argument is a
4725 '|'-separated list of channel names in order as they are given as
4726 additional stream inputs for filter. This also specify number of
4727 input streams. Number of input streams must be not less than number
4728 of channels in first stream plus one.
4729 gain
4731 Set gain applied to audio. Value is in dB. Default is 0.
4732 type
4734 Set processing type. Can be time or freq. time is processing audio
4735 in time domain which is slow. freq is processing audio in
4736 frequency domain which is fast. Default is freq.
4737 lfe Set custom gain for LFE channels. Value is in dB. Default is 0.
4739 size
4741 Set size of frame in number of samples which will be processed at
4742 once. Default value is 1024. Allowed range is from 1024 to 96000.
4743 hrir
4745 Set format of hrir stream. Default value is stereo. Alternative
4746 value is multich. If value is set to stereo, number of additional
4747 streams should be greater or equal to number of input channels in
4748 first input stream. Also each additional stream should have stereo
4749 number of channels. If value is set to multich, number of
4750 additional streams should be exactly one. Also number of input
4751 channels of additional stream should be equal or greater than twice
4752 number of channels of first input stream.
4753 Examples
4755 • Full example using wav files as coefficients with amovie filters
4757 for 7.1 downmix, each amovie filter use stereo file with IR
4758 coefficients as input. The files give coefficients for each
4759 position of virtual loudspeaker:
4760 ffmpeg -i input.wav
4762 -filter_complex "amovie=azi_270_ele_0_DFC.wav[sr];amovie=azi_90_ele_0_DFC.wav[sl];amovie=azi_225_ele_0_DFC.wav[br];amovie=azi_135_ele_0_DFC.wav[bl];amovie=azi_0_ele_0_DFC.wav,asplit[fc][lfe];amovie=azi_35_ele_0_DFC.wav[fl];amovie=azi_325_ele_0_DFC.wav[fr];[0:a][fl][fr][fc][lfe][bl][br][sl][sr]headphone=FL|FR|FC|LFE|BL|BR|SL|SR"
4763 output.wav
4764 • Full example using wav files as coefficients with amovie filters
4766 for 7.1 downmix, but now in multich hrir format.
4767 ffmpeg -i input.wav -filter_complex "amovie=minp.wav[hrirs];[0:a][hrirs]headphone=map=FL|FR|FC|LFE|BL|BR|SL|SR:hrir=multich"
4769 output.wav
4770 highpass
4772 Apply a high-pass filter with 3dB point frequency. The filter can be
4773 either single-pole, or double-pole (the default). The filter roll off
4774 at 6dB per pole per octave (20dB per pole per decade).
4775 The filter accepts the following options:
4777 frequency, f
4779 Set frequency in Hz. Default is 3000.
4780 poles, p
4782 Set number of poles. Default is 2.
4783 width_type, t
4785 Set method to specify band-width of filter.
4786 h Hz
4788 q Q-Factor
4790 o octave
4792 s slope
4794 k kHz
4796 width, w
4798 Specify the band-width of a filter in width_type units. Applies
4799 only to double-pole filter. The default is 0.707q and gives a
4800 Butterworth response.
4801 mix, m
4803 How much to use filtered signal in output. Default is 1. Range is
4804 between 0 and 1.
4805 channels, c
4807 Specify which channels to filter, by default all available are
4808 filtered.
4809 normalize, n
4811 Normalize biquad coefficients, by default is disabled. Enabling it
4812 will normalize magnitude response at DC to 0dB.
4813 transform, a
4815 Set transform type of IIR filter.
4816 di
4818 dii
4819 tdi
4820 tdii
4821 latt
4822 svf
4823 zdf
4824 precision, r
4825 Set precison of filtering.
4826 auto
4828 Pick automatic sample format depending on surround filters.
4829 s16 Always use signed 16-bit.
4831 s32 Always use signed 32-bit.
4833 f32 Always use float 32-bit.
4835 f64 Always use float 64-bit.
4837 block_size, b
4839 Set block size used for reverse IIR processing. If this value is
4840 set to high enough value (higher than impulse response length
4841 truncated when reaches near zero values) filtering will become
4842 linear phase otherwise if not big enough it will just produce nasty
4843 artifacts.
4844 Note that filter delay will be exactly this many samples when set
4846 to non-zero value.
4847 Commands
4849 This filter supports the following commands:
4851 frequency, f
4853 Change highpass frequency. Syntax for the command is : "frequency"
4854 width_type, t
4856 Change highpass width_type. Syntax for the command is :
4857 "width_type"
4858 width, w
4860 Change highpass width. Syntax for the command is : "width"
4861 mix, m
4863 Change highpass mix. Syntax for the command is : "mix"
4864 join
4866 Join multiple input streams into one multi-channel stream.
4867 It accepts the following parameters:
4869 inputs
4871 The number of input streams. It defaults to 2.
4872 channel_layout
4874 The desired output channel layout. It defaults to stereo.
4875 map Map channels from inputs to output. The argument is a '|'-separated
4877 list of mappings, each in the "input_idx.in_channel-out_channel"
4878 form. input_idx is the 0-based index of the input stream.
4879 in_channel can be either the name of the input channel (e.g. FL for
4880 front left) or its index in the specified input stream. out_channel
4881 is the name of the output channel.
4882 The filter will attempt to guess the mappings when they are not
4884 specified explicitly. It does so by first trying to find an unused
4885 matching input channel and if that fails it picks the first unused
4886 input channel.
4887 Join 3 inputs (with properly set channel layouts):
4889 ffmpeg -i INPUT1 -i INPUT2 -i INPUT3 -filter_complex join=inputs=3 OUTPUT
4891 Build a 5.1 output from 6 single-channel streams:
4893 ffmpeg -i fl -i fr -i fc -i sl -i sr -i lfe -filter_complex
4895 'join=inputs=6:channel_layout=5.1:map=0.0-FL|1.0-FR|2.0-FC|3.0-SL|4.0-SR|5.0-LFE'
4896 out
4897 ladspa
4899 Load a LADSPA (Linux Audio Developer's Simple Plugin API) plugin.
4900 To enable compilation of this filter you need to configure FFmpeg with
4902 "--enable-ladspa".
4903 file, f
4905 Specifies the name of LADSPA plugin library to load. If the
4906 environment variable LADSPA_PATH is defined, the LADSPA plugin is
4907 searched in each one of the directories specified by the colon
4908 separated list in LADSPA_PATH, otherwise in the standard LADSPA
4909 paths, which are in this order: HOME/.ladspa/lib/,
4910 /usr/local/lib/ladspa/, /usr/lib/ladspa/.
4911 plugin, p
4913 Specifies the plugin within the library. Some libraries contain
4914 only one plugin, but others contain many of them. If this is not
4915 set filter will list all available plugins within the specified
4916 library.
4917 controls, c
4919 Set the '|' separated list of controls which are zero or more
4920 floating point values that determine the behavior of the loaded
4921 plugin (for example delay, threshold or gain). Controls need to be
4922 defined using the following syntax:
4923 c0=value0|c1=value1|c2=value2|..., where valuei is the value set on
4924 the i-th control. Alternatively they can be also defined using the
4925 following syntax: value0|value1|value2|..., where valuei is the
4926 value set on the i-th control. If controls is set to "help", all
4927 available controls and their valid ranges are printed.
4928 sample_rate, s
4930 Specify the sample rate, default to 44100. Only used if plugin have
4931 zero inputs.
4932 nb_samples, n
4934 Set the number of samples per channel per each output frame,
4935 default is 1024. Only used if plugin have zero inputs.
4936 duration, d
4938 Set the minimum duration of the sourced audio. See the Time
4939 duration section in the ffmpeg-utils(1) manual for the accepted
4940 syntax. Note that the resulting duration may be greater than the
4941 specified duration, as the generated audio is always cut at the end
4942 of a complete frame. If not specified, or the expressed duration
4943 is negative, the audio is supposed to be generated forever. Only
4944 used if plugin have zero inputs.
4945 latency, l
4947 Enable latency compensation, by default is disabled. Only used if
4948 plugin have inputs.
4949 Examples
4951 • List all available plugins within amp (LADSPA example plugin)
4953 library:
4954 ladspa=file=amp
4956 • List all available controls and their valid ranges for "vcf_notch"
4958 plugin from "VCF" library:
4959 ladspa=f=vcf:p=vcf_notch:c=help
4961 • Simulate low quality audio equipment using "Computer Music Toolkit"
4963 (CMT) plugin library:
4964 ladspa=file=cmt:plugin=lofi:controls=c0=22|c1=12|c2=12
4966 • Add reverberation to the audio using TAP-plugins (Tom's Audio
4968 Processing plugins):
4969 ladspa=file=tap_reverb:tap_reverb
4971 • Generate white noise, with 0.2 amplitude:
4973 ladspa=file=cmt:noise_source_white:c=c0=.2
4975 • Generate 20 bpm clicks using plugin "C* Click - Metronome" from the
4977 "C* Audio Plugin Suite" (CAPS) library:
4978 ladspa=file=caps:Click:c=c1=20'
4980 • Apply "C* Eq10X2 - Stereo 10-band equaliser" effect:
4982 ladspa=caps:Eq10X2:c=c0=-48|c9=-24|c3=12|c4=2
4984 • Increase volume by 20dB using fast lookahead limiter from Steve
4986 Harris "SWH Plugins" collection:
4987 ladspa=fast_lookahead_limiter_1913:fastLookaheadLimiter:20|0|2
4989 • Attenuate low frequencies using Multiband EQ from Steve Harris "SWH
4991 Plugins" collection:
4992 ladspa=mbeq_1197:mbeq:-24|-24|-24|0|0|0|0|0|0|0|0|0|0|0|0
4994 • Reduce stereo image using "Narrower" from the "C* Audio Plugin
4996 Suite" (CAPS) library:
4997 ladspa=caps:Narrower
4999 • Another white noise, now using "C* Audio Plugin Suite" (CAPS)
5001 library:
5002 ladspa=caps:White:.2
5004 • Some fractal noise, using "C* Audio Plugin Suite" (CAPS) library:
5006 ladspa=caps:Fractal:c=c1=1
5008 • Dynamic volume normalization using "VLevel" plugin:
5010 ladspa=vlevel-ladspa:vlevel_mono
5012 Commands
5014 This filter supports the following commands:
5016 cN Modify the N-th control value.
5018 If the specified value is not valid, it is ignored and prior one is
5020 kept.
5021 loudnorm
5023 EBU R128 loudness normalization. Includes both dynamic and linear
5024 normalization modes. Support for both single pass (livestreams, files)
5025 and double pass (files) modes. This algorithm can target IL, LRA, and
5026 maximum true peak. In dynamic mode, to accurately detect true peaks,
5027 the audio stream will be upsampled to 192 kHz. Use the "-ar" option or
5028 "aresample" filter to explicitly set an output sample rate.
5029 The filter accepts the following options:
5031 I, i
5033 Set integrated loudness target. Range is -70.0 - -5.0. Default
5034 value is -24.0.
5035 LRA, lra
5037 Set loudness range target. Range is 1.0 - 50.0. Default value is
5038 7.0.
5039 TP, tp
5041 Set maximum true peak. Range is -9.0 - +0.0. Default value is
5042 -2.0.
5043 measured_I, measured_i
5045 Measured IL of input file. Range is -99.0 - +0.0.
5046 measured_LRA, measured_lra
5048 Measured LRA of input file. Range is 0.0 - 99.0.
5049 measured_TP, measured_tp
5051 Measured true peak of input file. Range is -99.0 - +99.0.
5052 measured_thresh
5054 Measured threshold of input file. Range is -99.0 - +0.0.
5055 offset
5057 Set offset gain. Gain is applied before the true-peak limiter.
5058 Range is -99.0 - +99.0. Default is +0.0.
5059 linear
5061 Normalize by linearly scaling the source audio. "measured_I",
5062 "measured_LRA", "measured_TP", and "measured_thresh" must all be
5063 specified. Target LRA shouldn't be lower than source LRA and the
5064 change in integrated loudness shouldn't result in a true peak which
5065 exceeds the target TP. If any of these conditions aren't met,
5066 normalization mode will revert to dynamic. Options are "true" or
5067 "false". Default is "true".
5068 dual_mono
5070 Treat mono input files as "dual-mono". If a mono file is intended
5071 for playback on a stereo system, its EBU R128 measurement will be
5072 perceptually incorrect. If set to "true", this option will
5073 compensate for this effect. Multi-channel input files are not
5074 affected by this option. Options are true or false. Default is
5075 false.
5076 print_format
5078 Set print format for stats. Options are summary, json, or none.
5079 Default value is none.
5080 lowpass
5082 Apply a low-pass filter with 3dB point frequency. The filter can be
5083 either single-pole or double-pole (the default). The filter roll off
5084 at 6dB per pole per octave (20dB per pole per decade).
5085 The filter accepts the following options:
5087 frequency, f
5089 Set frequency in Hz. Default is 500.
5090 poles, p
5092 Set number of poles. Default is 2.
5093 width_type, t
5095 Set method to specify band-width of filter.
5096 h Hz
5098 q Q-Factor
5100 o octave
5102 s slope
5104 k kHz
5106 width, w
5108 Specify the band-width of a filter in width_type units. Applies
5109 only to double-pole filter. The default is 0.707q and gives a
5110 Butterworth response.
5111 mix, m
5113 How much to use filtered signal in output. Default is 1. Range is
5114 between 0 and 1.
5115 channels, c
5117 Specify which channels to filter, by default all available are
5118 filtered.
5119 normalize, n
5121 Normalize biquad coefficients, by default is disabled. Enabling it
5122 will normalize magnitude response at DC to 0dB.
5123 transform, a
5125 Set transform type of IIR filter.
5126 di
5128 dii
5129 tdi
5130 tdii
5131 latt
5132 svf
5133 zdf
5134 precision, r
5135 Set precison of filtering.
5136 auto
5138 Pick automatic sample format depending on surround filters.
5139 s16 Always use signed 16-bit.
5141 s32 Always use signed 32-bit.
5143 f32 Always use float 32-bit.
5145 f64 Always use float 64-bit.
5147 block_size, b
5149 Set block size used for reverse IIR processing. If this value is
5150 set to high enough value (higher than impulse response length
5151 truncated when reaches near zero values) filtering will become
5152 linear phase otherwise if not big enough it will just produce nasty
5153 artifacts.
5154 Note that filter delay will be exactly this many samples when set
5156 to non-zero value.
5157 Examples
5159 • Lowpass only LFE channel, it LFE is not present it does nothing:
5161 lowpass=c=LFE
5163 Commands
5165 This filter supports the following commands:
5167 frequency, f
5169 Change lowpass frequency. Syntax for the command is : "frequency"
5170 width_type, t
5172 Change lowpass width_type. Syntax for the command is :
5173 "width_type"
5174 width, w
5176 Change lowpass width. Syntax for the command is : "width"
5177 mix, m
5179 Change lowpass mix. Syntax for the command is : "mix"
5180 lv2
5182 Load a LV2 (LADSPA Version 2) plugin.
5183 To enable compilation of this filter you need to configure FFmpeg with
5185 "--enable-lv2".
5186 plugin, p
5188 Specifies the plugin URI. You may need to escape ':'.
5189 controls, c
5191 Set the '|' separated list of controls which are zero or more
5192 floating point values that determine the behavior of the loaded
5193 plugin (for example delay, threshold or gain). If controls is set
5194 to "help", all available controls and their valid ranges are
5195 printed.
5196 sample_rate, s
5198 Specify the sample rate, default to 44100. Only used if plugin have
5199 zero inputs.
5200 nb_samples, n
5202 Set the number of samples per channel per each output frame,
5203 default is 1024. Only used if plugin have zero inputs.
5204 duration, d
5206 Set the minimum duration of the sourced audio. See the Time
5207 duration section in the ffmpeg-utils(1) manual for the accepted
5208 syntax. Note that the resulting duration may be greater than the
5209 specified duration, as the generated audio is always cut at the end
5210 of a complete frame. If not specified, or the expressed duration
5211 is negative, the audio is supposed to be generated forever. Only
5212 used if plugin have zero inputs.
5213 Examples
5215 • Apply bass enhancer plugin from Calf:
5217 lv2=p=http\\\\://calf.sourceforge.net/plugins/BassEnhancer:c=amount=2
5219 • Apply vinyl plugin from Calf:
5221 lv2=p=http\\\\://calf.sourceforge.net/plugins/Vinyl:c=drone=0.2|aging=0.5
5223 • Apply bit crusher plugin from ArtyFX:
5225 lv2=p=http\\\\://www.openavproductions.com/artyfx#bitta:c=crush=0.3
5227 Commands
5229 This filter supports all options that are exported by plugin as
5231 commands.
5232 mcompand
5234 Multiband Compress or expand the audio's dynamic range.
5235 The input audio is divided into bands using 4th order Linkwitz-Riley
5237 IIRs. This is akin to the crossover of a loudspeaker, and results in
5238 flat frequency response when absent compander action.
5239 It accepts the following parameters:
5241 args
5243 This option syntax is: attack,decay,[attack,decay..] soft-knee
5244 points crossover_frequency [delay [initial_volume [gain]]] |
5245 attack,decay ... For explanation of each item refer to compand
5246 filter documentation.
5247 pan
5249 Mix channels with specific gain levels. The filter accepts the output
5250 channel layout followed by a set of channels definitions.
5251 This filter is also designed to efficiently remap the channels of an
5253 audio stream.
5254 The filter accepts parameters of the form: "l|outdef|outdef|..."
5256 l output channel layout or number of channels
5258 outdef
5260 output channel specification, of the form:
5261 "out_name=[gain*]in_name[(+-)[gain*]in_name...]"
5262 out_name
5264 output channel to define, either a channel name (FL, FR, etc.) or a
5265 channel number (c0, c1, etc.)
5266 gain
5268 multiplicative coefficient for the channel, 1 leaving the volume
5269 unchanged
5270 in_name
5272 input channel to use, see out_name for details; it is not possible
5273 to mix named and numbered input channels
5274 If the `=' in a channel specification is replaced by `<', then the
5276 gains for that specification will be renormalized so that the total is
5277 1, thus avoiding clipping noise.
5278 Mixing examples
5280 For example, if you want to down-mix from stereo to mono, but with a
5282 bigger factor for the left channel:
5283 pan=1c|c0=0.9*c0+0.1*c1
5285 A customized down-mix to stereo that works automatically for 3-, 4-, 5-
5287 and 7-channels surround:
5288 pan=stereo| FL < FL + 0.5*FC + 0.6*BL + 0.6*SL | FR < FR + 0.5*FC + 0.6*BR + 0.6*SR
5290 Note that ffmpeg integrates a default down-mix (and up-mix) system that
5292 should be preferred (see "-ac" option) unless you have very specific
5293 needs.
5294 Remapping examples
5296 The channel remapping will be effective if, and only if:
5298 *<gain coefficients are zeroes or ones,>
5300 *<only one input per channel output,>
5301 If all these conditions are satisfied, the filter will notify the user
5303 ("Pure channel mapping detected"), and use an optimized and lossless
5304 method to do the remapping.
5305 For example, if you have a 5.1 source and want a stereo audio stream by
5307 dropping the extra channels:
5308 pan="stereo| c0=FL | c1=FR"
5310 Given the same source, you can also switch front left and front right
5312 channels and keep the input channel layout:
5313 pan="5.1| c0=c1 | c1=c0 | c2=c2 | c3=c3 | c4=c4 | c5=c5"
5315 If the input is a stereo audio stream, you can mute the front left
5317 channel (and still keep the stereo channel layout) with:
5318 pan="stereo|c1=c1"
5320 Still with a stereo audio stream input, you can copy the right channel
5322 in both front left and right:
5323 pan="stereo| c0=FR | c1=FR"
5325 replaygain
5327 ReplayGain scanner filter. This filter takes an audio stream as an
5328 input and outputs it unchanged. At end of filtering it displays
5329 "track_gain" and "track_peak".
5330 resample
5332 Convert the audio sample format, sample rate and channel layout. It is
5333 not meant to be used directly.
5334 rubberband
5336 Apply time-stretching and pitch-shifting with librubberband.
5337 To enable compilation of this filter, you need to configure FFmpeg with
5339 "--enable-librubberband".
5340 The filter accepts the following options:
5342 tempo
5344 Set tempo scale factor.
5345 pitch
5347 Set pitch scale factor.
5348 transients
5350 Set transients detector. Possible values are:
5351 crisp
5353 mixed
5354 smooth
5355 detector
5356 Set detector. Possible values are:
5357 compound
5359 percussive
5360 soft
5361 phase
5362 Set phase. Possible values are:
5363 laminar
5365 independent
5366 window
5367 Set processing window size. Possible values are:
5368 standard
5370 short
5371 long
5372 smoothing
5373 Set smoothing. Possible values are:
5374 off
5376 on
5377 formant
5378 Enable formant preservation when shift pitching. Possible values
5379 are:
5380 shifted
5382 preserved
5383 pitchq
5384 Set pitch quality. Possible values are:
5385 quality
5387 speed
5388 consistency
5389 channels
5390 Set channels. Possible values are:
5391 apart
5393 together
5394 Commands
5396 This filter supports the following commands:
5398 tempo
5400 Change filter tempo scale factor. Syntax for the command is :
5401 "tempo"
5402 pitch
5404 Change filter pitch scale factor. Syntax for the command is :
5405 "pitch"
5406 sidechaincompress
5408 This filter acts like normal compressor but has the ability to compress
5409 detected signal using second input signal. It needs two input streams
5410 and returns one output stream. First input stream will be processed
5411 depending on second stream signal. The filtered signal then can be
5412 filtered with other filters in later stages of processing. See pan and
5413 amerge filter.
5414 The filter accepts the following options:
5416 level_in
5418 Set input gain. Default is 1. Range is between 0.015625 and 64.
5419 mode
5421 Set mode of compressor operation. Can be "upward" or "downward".
5422 Default is "downward".
5423 threshold
5425 If a signal of second stream raises above this level it will affect
5426 the gain reduction of first stream. By default is 0.125. Range is
5427 between 0.00097563 and 1.
5428 ratio
5430 Set a ratio about which the signal is reduced. 1:2 means that if
5431 the level raised 4dB above the threshold, it will be only 2dB above
5432 after the reduction. Default is 2. Range is between 1 and 20.
5433 attack
5435 Amount of milliseconds the signal has to rise above the threshold
5436 before gain reduction starts. Default is 20. Range is between 0.01
5437 and 2000.
5438 release
5440 Amount of milliseconds the signal has to fall below the threshold
5441 before reduction is decreased again. Default is 250. Range is
5442 between 0.01 and 9000.
5443 makeup
5445 Set the amount by how much signal will be amplified after
5446 processing. Default is 1. Range is from 1 to 64.
5447 knee
5449 Curve the sharp knee around the threshold to enter gain reduction
5450 more softly. Default is 2.82843. Range is between 1 and 8.
5451 link
5453 Choose if the "average" level between all channels of side-chain
5454 stream or the louder("maximum") channel of side-chain stream
5455 affects the reduction. Default is "average".
5456 detection
5458 Should the exact signal be taken in case of "peak" or an RMS one in
5459 case of "rms". Default is "rms" which is mainly smoother.
5460 level_sc
5462 Set sidechain gain. Default is 1. Range is between 0.015625 and 64.
5463 mix How much to use compressed signal in output. Default is 1. Range
5465 is between 0 and 1.
5466 Commands
5468 This filter supports the all above options as commands.
5470 Examples
5472 • Full ffmpeg example taking 2 audio inputs, 1st input to be
5474 compressed depending on the signal of 2nd input and later
5475 compressed signal to be merged with 2nd input:
5476 ffmpeg -i main.flac -i sidechain.flac -filter_complex "[1:a]asplit=2[sc][mix];[0:a][sc]sidechaincompress[compr];[compr][mix]amerge"
5478 sidechaingate
5480 A sidechain gate acts like a normal (wideband) gate but has the ability
5481 to filter the detected signal before sending it to the gain reduction
5482 stage. Normally a gate uses the full range signal to detect a level
5483 above the threshold. For example: If you cut all lower frequencies
5484 from your sidechain signal the gate will decrease the volume of your
5485 track only if not enough highs appear. With this technique you are able
5486 to reduce the resonation of a natural drum or remove "rumbling" of
5487 muted strokes from a heavily distorted guitar. It needs two input
5488 streams and returns one output stream. First input stream will be
5489 processed depending on second stream signal.
5490 The filter accepts the following options:
5492 level_in
5494 Set input level before filtering. Default is 1. Allowed range is
5495 from 0.015625 to 64.
5496 mode
5498 Set the mode of operation. Can be "upward" or "downward". Default
5499 is "downward". If set to "upward" mode, higher parts of signal will
5500 be amplified, expanding dynamic range in upward direction.
5501 Otherwise, in case of "downward" lower parts of signal will be
5502 reduced.
5503 range
5505 Set the level of gain reduction when the signal is below the
5506 threshold. Default is 0.06125. Allowed range is from 0 to 1.
5507 Setting this to 0 disables reduction and then filter behaves like
5508 expander.
5509 threshold
5511 If a signal rises above this level the gain reduction is released.
5512 Default is 0.125. Allowed range is from 0 to 1.
5513 ratio
5515 Set a ratio about which the signal is reduced. Default is 2.
5516 Allowed range is from 1 to 9000.
5517 attack
5519 Amount of milliseconds the signal has to rise above the threshold
5520 before gain reduction stops. Default is 20 milliseconds. Allowed
5521 range is from 0.01 to 9000.
5522 release
5524 Amount of milliseconds the signal has to fall below the threshold
5525 before the reduction is increased again. Default is 250
5526 milliseconds. Allowed range is from 0.01 to 9000.
5527 makeup
5529 Set amount of amplification of signal after processing. Default is
5530 1. Allowed range is from 1 to 64.
5531 knee
5533 Curve the sharp knee around the threshold to enter gain reduction
5534 more softly. Default is 2.828427125. Allowed range is from 1 to 8.
5535 detection
5537 Choose if exact signal should be taken for detection or an RMS like
5538 one. Default is rms. Can be peak or rms.
5539 link
5541 Choose if the average level between all channels or the louder
5542 channel affects the reduction. Default is average. Can be average
5543 or maximum.
5544 level_sc
5546 Set sidechain gain. Default is 1. Range is from 0.015625 to 64.
5547 Commands
5549 This filter supports the all above options as commands.
5551 silencedetect
5553 Detect silence in an audio stream.
5554 This filter logs a message when it detects that the input audio volume
5556 is less or equal to a noise tolerance value for a duration greater or
5557 equal to the minimum detected noise duration.
5558 The printed times and duration are expressed in seconds. The
5560 "lavfi.silence_start" or "lavfi.silence_start.X" metadata key is set on
5561 the first frame whose timestamp equals or exceeds the detection
5562 duration and it contains the timestamp of the first frame of the
5563 silence.
5564 The "lavfi.silence_duration" or "lavfi.silence_duration.X" and
5566 "lavfi.silence_end" or "lavfi.silence_end.X" metadata keys are set on
5567 the first frame after the silence. If mono is enabled, and each channel
5568 is evaluated separately, the ".X" suffixed keys are used, and "X"
5569 corresponds to the channel number.
5570 The filter accepts the following options:
5572 noise, n
5574 Set noise tolerance. Can be specified in dB (in case "dB" is
5575 appended to the specified value) or amplitude ratio. Default is
5576 -60dB, or 0.001.
5577 duration, d
5579 Set silence duration until notification (default is 2 seconds). See
5580 the Time duration section in the ffmpeg-utils(1) manual for the
5581 accepted syntax.
5582 mono, m
5584 Process each channel separately, instead of combined. By default is
5585 disabled.
5586 Examples
5588 • Detect 5 seconds of silence with -50dB noise tolerance:
5590 silencedetect=n=-50dB:d=5
5592 • Complete example with ffmpeg to detect silence with 0.0001 noise
5594 tolerance in silence.mp3:
5595 ffmpeg -i silence.mp3 -af silencedetect=noise=0.0001 -f null -
5597 silenceremove
5599 Remove silence from the beginning, middle or end of the audio.
5600 The filter accepts the following options:
5602 start_periods
5604 This value is used to indicate if audio should be trimmed at
5605 beginning of the audio. A value of zero indicates no silence should
5606 be trimmed from the beginning. When specifying a non-zero value, it
5607 trims audio up until it finds non-silence. Normally, when trimming
5608 silence from beginning of audio the start_periods will be 1 but it
5609 can be increased to higher values to trim all audio up to specific
5610 count of non-silence periods. Default value is 0.
5611 start_duration
5613 Specify the amount of time that non-silence must be detected before
5614 it stops trimming audio. By increasing the duration, bursts of
5615 noises can be treated as silence and trimmed off. Default value is
5616 0.
5617 start_threshold
5619 This indicates what sample value should be treated as silence. For
5620 digital audio, a value of 0 may be fine but for audio recorded from
5621 analog, you may wish to increase the value to account for
5622 background noise. Can be specified in dB (in case "dB" is appended
5623 to the specified value) or amplitude ratio. Default value is 0.
5624 start_silence
5626 Specify max duration of silence at beginning that will be kept
5627 after trimming. Default is 0, which is equal to trimming all
5628 samples detected as silence.
5629 start_mode
5631 Specify mode of detection of silence end in start of multi-channel
5632 audio. Can be any or all. Default is any. With any, any sample
5633 that is detected as non-silence will cause stopped trimming of
5634 silence. With all, only if all channels are detected as non-
5635 silence will cause stopped trimming of silence.
5636 stop_periods
5638 Set the count for trimming silence from the end of audio. To
5639 remove silence from the middle of a file, specify a stop_periods
5640 that is negative. This value is then treated as a positive value
5641 and is used to indicate the effect should restart processing as
5642 specified by start_periods, making it suitable for removing periods
5643 of silence in the middle of the audio. Default value is 0.
5644 stop_duration
5646 Specify a duration of silence that must exist before audio is not
5647 copied any more. By specifying a higher duration, silence that is
5648 wanted can be left in the audio. Default value is 0.
5649 stop_threshold
5651 This is the same as start_threshold but for trimming silence from
5652 the end of audio. Can be specified in dB (in case "dB" is appended
5653 to the specified value) or amplitude ratio. Default value is 0.
5654 stop_silence
5656 Specify max duration of silence at end that will be kept after
5657 trimming. Default is 0, which is equal to trimming all samples
5658 detected as silence.
5659 stop_mode
5661 Specify mode of detection of silence start in end of multi-channel
5662 audio. Can be any or all. Default is any. With any, any sample
5663 that is detected as non-silence will cause stopped trimming of
5664 silence. With all, only if all channels are detected as non-
5665 silence will cause stopped trimming of silence.
5666 detection
5668 Set how is silence detected. Can be "rms" or "peak". Second is
5669 faster and works better with digital silence which is exactly 0.
5670 Default value is "rms".
5671 window
5673 Set duration in number of seconds used to calculate size of window
5674 in number of samples for detecting silence. Default value is 0.02.
5675 Allowed range is from 0 to 10.
5676 Examples
5678 • The following example shows how this filter can be used to start a
5680 recording that does not contain the delay at the start which
5681 usually occurs between pressing the record button and the start of
5682 the performance:
5683 silenceremove=start_periods=1:start_duration=5:start_threshold=0.02
5685 • Trim all silence encountered from beginning to end where there is
5687 more than 1 second of silence in audio:
5688 silenceremove=stop_periods=-1:stop_duration=1:stop_threshold=-90dB
5690 • Trim all digital silence samples, using peak detection, from
5692 beginning to end where there is more than 0 samples of digital
5693 silence in audio and digital silence is detected in all channels at
5694 same positions in stream:
5695 silenceremove=window=0:detection=peak:stop_mode=all:start_mode=all:stop_periods=-1:stop_threshold=0
5697 sofalizer
5699 SOFAlizer uses head-related transfer functions (HRTFs) to create
5700 virtual loudspeakers around the user for binaural listening via
5701 headphones (audio formats up to 9 channels supported). The HRTFs are
5702 stored in SOFA files (see <http://www.sofacoustics.org/> for a
5703 database). SOFAlizer is developed at the Acoustics Research Institute
5704 (ARI) of the Austrian Academy of Sciences.
5705 To enable compilation of this filter you need to configure FFmpeg with
5707 "--enable-libmysofa".
5708 The filter accepts the following options:
5710 sofa
5712 Set the SOFA file used for rendering.
5713 gain
5715 Set gain applied to audio. Value is in dB. Default is 0.
5716 rotation
5718 Set rotation of virtual loudspeakers in deg. Default is 0.
5719 elevation
5721 Set elevation of virtual speakers in deg. Default is 0.
5722 radius
5724 Set distance in meters between loudspeakers and the listener with
5725 near-field HRTFs. Default is 1.
5726 type
5728 Set processing type. Can be time or freq. time is processing audio
5729 in time domain which is slow. freq is processing audio in
5730 frequency domain which is fast. Default is freq.
5731 speakers
5733 Set custom positions of virtual loudspeakers. Syntax for this
5734 option is: <CH> <AZIM> <ELEV>[|<CH> <AZIM> <ELEV>|...]. Each
5735 virtual loudspeaker is described with short channel name following
5736 with azimuth and elevation in degrees. Each virtual loudspeaker
5737 description is separated by '|'. For example to override front
5738 left and front right channel positions use: 'speakers=FL 45 15|FR
5739 345 15'. Descriptions with unrecognised channel names are ignored.
5740 lfegain
5742 Set custom gain for LFE channels. Value is in dB. Default is 0.
5743 framesize
5745 Set custom frame size in number of samples. Default is 1024.
5746 Allowed range is from 1024 to 96000. Only used if option type is
5747 set to freq.
5748 normalize
5750 Should all IRs be normalized upon importing SOFA file. By default
5751 is enabled.
5752 interpolate
5754 Should nearest IRs be interpolated with neighbor IRs if exact
5755 position does not match. By default is disabled.
5756 minphase
5758 Minphase all IRs upon loading of SOFA file. By default is disabled.
5759 anglestep
5761 Set neighbor search angle step. Only used if option interpolate is
5762 enabled.
5763 radstep
5765 Set neighbor search radius step. Only used if option interpolate is
5766 enabled.
5767 Examples
5769 • Using ClubFritz6 sofa file:
5771 sofalizer=sofa=/path/to/ClubFritz6.sofa:type=freq:radius=1
5773 • Using ClubFritz12 sofa file and bigger radius with small rotation:
5775 sofalizer=sofa=/path/to/ClubFritz12.sofa:type=freq:radius=2:rotation=5
5777 • Similar as above but with custom speaker positions for front left,
5779 front right, back left and back right and also with custom gain:
5780 "sofalizer=sofa=/path/to/ClubFritz6.sofa:type=freq:radius=2:speakers=FL 45|FR 315|BL 135|BR 225:gain=28"
5782 speechnorm
5784 Speech Normalizer.
5785 This filter expands or compresses each half-cycle of audio samples
5787 (local set of samples all above or all below zero and between two
5788 nearest zero crossings) depending on threshold value, so audio reaches
5789 target peak value under conditions controlled by below options.
5790 The filter accepts the following options:
5792 peak, p
5794 Set the expansion target peak value. This specifies the highest
5795 allowed absolute amplitude level for the normalized audio input.
5796 Default value is 0.95. Allowed range is from 0.0 to 1.0.
5797 expansion, e
5799 Set the maximum expansion factor. Allowed range is from 1.0 to
5800 50.0. Default value is 2.0. This option controls maximum local
5801 half-cycle of samples expansion. The maximum expansion would be
5802 such that local peak value reaches target peak value but never to
5803 surpass it and that ratio between new and previous peak value does
5804 not surpass this option value.
5805 compression, c
5807 Set the maximum compression factor. Allowed range is from 1.0 to
5808 50.0. Default value is 2.0. This option controls maximum local
5809 half-cycle of samples compression. This option is used only if
5810 threshold option is set to value greater than 0.0, then in such
5811 cases when local peak is lower or same as value set by threshold
5812 all samples belonging to that peak's half-cycle will be compressed
5813 by current compression factor.
5814 threshold, t
5816 Set the threshold value. Default value is 0.0. Allowed range is
5817 from 0.0 to 1.0. This option specifies which half-cycles of
5818 samples will be compressed and which will be expanded. Any half-
5819 cycle samples with their local peak value below or same as this
5820 option value will be compressed by current compression factor,
5821 otherwise, if greater than threshold value they will be expanded
5822 with expansion factor so that it could reach peak target value but
5823 never surpass it.
5824 raise, r
5826 Set the expansion raising amount per each half-cycle of samples.
5827 Default value is 0.001. Allowed range is from 0.0 to 1.0. This
5828 controls how fast expansion factor is raised per each new half-
5829 cycle until it reaches expansion value. Setting this options too
5830 high may lead to distortions.
5831 fall, f
5833 Set the compression raising amount per each half-cycle of samples.
5834 Default value is 0.001. Allowed range is from 0.0 to 1.0. This
5835 controls how fast compression factor is raised per each new half-
5836 cycle until it reaches compression value.
5837 channels, h
5839 Specify which channels to filter, by default all available channels
5840 are filtered.
5841 invert, i
5843 Enable inverted filtering, by default is disabled. This inverts
5844 interpretation of threshold option. When enabled any half-cycle of
5845 samples with their local peak value below or same as threshold
5846 option will be expanded otherwise it will be compressed.
5847 link, l
5849 Link channels when calculating gain applied to each filtered
5850 channel sample, by default is disabled. When disabled each
5851 filtered channel gain calculation is independent, otherwise when
5852 this option is enabled the minimum of all possible gains for each
5853 filtered channel is used.
5854 rms, m
5856 Set the expansion target RMS value. This specifies the highest
5857 allowed RMS level for the normalized audio input. Default value is
5858 0.0, thus disabled. Allowed range is from 0.0 to 1.0.
5859 Commands
5861 This filter supports the all above options as commands.
5863 Examples
5865 • Weak and slow amplification:
5867 speechnorm=e=3:r=0.00001:l=1
5869 • Moderate and slow amplification:
5871 speechnorm=e=6.25:r=0.00001:l=1
5873 • Strong and fast amplification:
5875 speechnorm=e=12.5:r=0.0001:l=1
5877 • Very strong and fast amplification:
5879 speechnorm=e=25:r=0.0001:l=1
5881 • Extreme and fast amplification:
5883 speechnorm=e=50:r=0.0001:l=1
5885 stereotools
5887 This filter has some handy utilities to manage stereo signals, for
5888 converting M/S stereo recordings to L/R signal while having control
5889 over the parameters or spreading the stereo image of master track.
5890 The filter accepts the following options:
5892 level_in
5894 Set input level before filtering for both channels. Defaults is 1.
5895 Allowed range is from 0.015625 to 64.
5896 level_out
5898 Set output level after filtering for both channels. Defaults is 1.
5899 Allowed range is from 0.015625 to 64.
5900 balance_in
5902 Set input balance between both channels. Default is 0. Allowed
5903 range is from -1 to 1.
5904 balance_out
5906 Set output balance between both channels. Default is 0. Allowed
5907 range is from -1 to 1.
5908 softclip
5910 Enable softclipping. Results in analog distortion instead of harsh
5911 digital 0dB clipping. Disabled by default.
5912 mutel
5914 Mute the left channel. Disabled by default.
5915 muter
5917 Mute the right channel. Disabled by default.
5918 phasel
5920 Change the phase of the left channel. Disabled by default.
5921 phaser
5923 Change the phase of the right channel. Disabled by default.
5924 mode
5926 Set stereo mode. Available values are:
5927 lr>lr
5929 Left/Right to Left/Right, this is default.
5930 lr>ms
5932 Left/Right to Mid/Side.
5933 ms>lr
5935 Mid/Side to Left/Right.
5936 lr>ll
5938 Left/Right to Left/Left.
5939 lr>rr
5941 Left/Right to Right/Right.
5942 lr>l+r
5944 Left/Right to Left + Right.
5945 lr>rl
5947 Left/Right to Right/Left.
5948 ms>ll
5950 Mid/Side to Left/Left.
5951 ms>rr
5953 Mid/Side to Right/Right.
5954 ms>rl
5956 Mid/Side to Right/Left.
5957 lr>l-r
5959 Left/Right to Left - Right.
5960 slev
5962 Set level of side signal. Default is 1. Allowed range is from
5963 0.015625 to 64.
5964 sbal
5966 Set balance of side signal. Default is 0. Allowed range is from -1
5967 to 1.
5968 mlev
5970 Set level of the middle signal. Default is 1. Allowed range is
5971 from 0.015625 to 64.
5972 mpan
5974 Set middle signal pan. Default is 0. Allowed range is from -1 to 1.
5975 base
5977 Set stereo base between mono and inversed channels. Default is 0.
5978 Allowed range is from -1 to 1.
5979 delay
5981 Set delay in milliseconds how much to delay left from right channel
5982 and vice versa. Default is 0. Allowed range is from -20 to 20.
5983 sclevel
5985 Set S/C level. Default is 1. Allowed range is from 1 to 100.
5986 phase
5988 Set the stereo phase in degrees. Default is 0. Allowed range is
5989 from 0 to 360.
5990 bmode_in, bmode_out
5992 Set balance mode for balance_in/balance_out option.
5993 Can be one of the following:
5995 balance
5997 Classic balance mode. Attenuate one channel at time. Gain is
5998 raised up to 1.
5999 amplitude
6001 Similar as classic mode above but gain is raised up to 2.
6002 power
6004 Equal power distribution, from -6dB to +6dB range.
6005 Commands
6007 This filter supports the all above options as commands.
6009 Examples
6011 • Apply karaoke like effect:
6013 stereotools=mlev=0.015625
6015 • Convert M/S signal to L/R:
6017 "stereotools=mode=ms>lr"
6019 stereowiden
6021 This filter enhance the stereo effect by suppressing signal common to
6022 both channels and by delaying the signal of left into right and vice
6023 versa, thereby widening the stereo effect.
6024 The filter accepts the following options:
6026 delay
6028 Time in milliseconds of the delay of left signal into right and
6029 vice versa. Default is 20 milliseconds.
6030 feedback
6032 Amount of gain in delayed signal into right and vice versa. Gives a
6033 delay effect of left signal in right output and vice versa which
6034 gives widening effect. Default is 0.3.
6035 crossfeed
6037 Cross feed of left into right with inverted phase. This helps in
6038 suppressing the mono. If the value is 1 it will cancel all the
6039 signal common to both channels. Default is 0.3.
6040 drymix
6042 Set level of input signal of original channel. Default is 0.8.
6043 Commands
6045 This filter supports the all above options except "delay" as commands.
6047 superequalizer
6049 Apply 18 band equalizer.
6050 The filter accepts the following options:
6052 1b Set 65Hz band gain.
6054 2b Set 92Hz band gain.
6056 3b Set 131Hz band gain.
6058 4b Set 185Hz band gain.
6060 5b Set 262Hz band gain.
6062 6b Set 370Hz band gain.
6064 7b Set 523Hz band gain.
6066 8b Set 740Hz band gain.
6068 9b Set 1047Hz band gain.
6070 10b Set 1480Hz band gain.
6072 11b Set 2093Hz band gain.
6074 12b Set 2960Hz band gain.
6076 13b Set 4186Hz band gain.
6078 14b Set 5920Hz band gain.
6080 15b Set 8372Hz band gain.
6082 16b Set 11840Hz band gain.
6084 17b Set 16744Hz band gain.
6086 18b Set 20000Hz band gain.
6088 surround
6090 Apply audio surround upmix filter.
6091 This filter allows to produce multichannel output from audio stream.
6093 The filter accepts the following options:
6095 chl_out
6097 Set output channel layout. By default, this is 5.1.
6098 See the Channel Layout section in the ffmpeg-utils(1) manual for
6100 the required syntax.
6101 chl_in
6103 Set input channel layout. By default, this is stereo.
6104 See the Channel Layout section in the ffmpeg-utils(1) manual for
6106 the required syntax.
6107 level_in
6109 Set input volume level. By default, this is 1.
6110 level_out
6112 Set output volume level. By default, this is 1.
6113 lfe Enable LFE channel output if output channel layout has it. By
6115 default, this is enabled.
6116 lfe_low
6118 Set LFE low cut off frequency. By default, this is 128 Hz.
6119 lfe_high
6121 Set LFE high cut off frequency. By default, this is 256 Hz.
6122 lfe_mode
6124 Set LFE mode, can be add or sub. Default is add. In add mode, LFE
6125 channel is created from input audio and added to output. In sub
6126 mode, LFE channel is created from input audio and added to output
6127 but also all non-LFE output channels are subtracted with output LFE
6128 channel.
6129 smooth
6131 Set temporal smoothness strength, used to gradually change factors
6132 when transforming stereo sound in time. Allowed range is from 0.0
6133 to 1.0. Useful to improve output quality with focus option values
6134 greater than 0.0. Default is 0.0. Only values inside this range
6135 and without edges are effective.
6136 angle
6138 Set angle of stereo surround transform, Allowed range is from 0 to
6139 360. Default is 90.
6140 focus
6142 Set focus of stereo surround transform, Allowed range is from -1 to
6143 1. Default is 0.
6144 fc_in
6146 Set front center input volume. By default, this is 1.
6147 fc_out
6149 Set front center output volume. By default, this is 1.
6150 fl_in
6152 Set front left input volume. By default, this is 1.
6153 fl_out
6155 Set front left output volume. By default, this is 1.
6156 fr_in
6158 Set front right input volume. By default, this is 1.
6159 fr_out
6161 Set front right output volume. By default, this is 1.
6162 sl_in
6164 Set side left input volume. By default, this is 1.
6165 sl_out
6167 Set side left output volume. By default, this is 1.
6168 sr_in
6170 Set side right input volume. By default, this is 1.
6171 sr_out
6173 Set side right output volume. By default, this is 1.
6174 bl_in
6176 Set back left input volume. By default, this is 1.
6177 bl_out
6179 Set back left output volume. By default, this is 1.
6180 br_in
6182 Set back right input volume. By default, this is 1.
6183 br_out
6185 Set back right output volume. By default, this is 1.
6186 bc_in
6188 Set back center input volume. By default, this is 1.
6189 bc_out
6191 Set back center output volume. By default, this is 1.
6192 lfe_in
6194 Set LFE input volume. By default, this is 1.
6195 lfe_out
6197 Set LFE output volume. By default, this is 1.
6198 allx
6200 Set spread usage of stereo image across X axis for all channels.
6201 Allowed range is from -1 to 15. By default this value is negative
6202 -1, and thus unused.
6203 ally
6205 Set spread usage of stereo image across Y axis for all channels.
6206 Allowed range is from -1 to 15. By default this value is negative
6207 -1, and thus unused.
6208 fcx, flx, frx, blx, brx, slx, srx, bcx
6210 Set spread usage of stereo image across X axis for each channel.
6211 Allowed range is from 0.06 to 15. By default this value is 0.5.
6212 fcy, fly, fry, bly, bry, sly, sry, bcy
6214 Set spread usage of stereo image across Y axis for each channel.
6215 Allowed range is from 0.06 to 15. By default this value is 0.5.
6216 win_size
6218 Set window size. Allowed range is from 1024 to 65536. Default size
6219 is 4096.
6220 win_func
6222 Set window function.
6223 It accepts the following values:
6225 rect
6227 bartlett
6228 hann, hanning
6229 hamming
6230 blackman
6231 welch
6232 flattop
6233 bharris
6234 bnuttall
6235 bhann
6236 sine
6237 nuttall
6238 lanczos
6239 gauss
6240 tukey
6241 dolph
6242 cauchy
6243 parzen
6244 poisson
6245 bohman
6246 kaiser
6247 Default is "hann".
6249 overlap
6251 Set window overlap. If set to 1, the recommended overlap for
6252 selected window function will be picked. Default is 0.5.
6253 tiltshelf
6255 Boost or cut the lower frequencies and cut or boost higher frequencies
6256 of the audio using a two-pole shelving filter with a response similar
6257 to that of a standard hi-fi's tone-controls. This is also known as
6258 shelving equalisation (EQ).
6259 The filter accepts the following options:
6261 gain, g
6263 Give the gain at 0 Hz. Its useful range is about -20 (for a large
6264 cut) to +20 (for a large boost). Beware of clipping when using a
6265 positive gain.
6266 frequency, f
6268 Set the filter's central frequency and so can be used to extend or
6269 reduce the frequency range to be boosted or cut. The default value
6270 is 3000 Hz.
6271 width_type, t
6273 Set method to specify band-width of filter.
6274 h Hz
6276 q Q-Factor
6278 o octave
6280 s slope
6282 k kHz
6284 width, w
6286 Determine how steep is the filter's shelf transition.
6287 poles, p
6289 Set number of poles. Default is 2.
6290 mix, m
6292 How much to use filtered signal in output. Default is 1. Range is
6293 between 0 and 1.
6294 channels, c
6296 Specify which channels to filter, by default all available are
6297 filtered.
6298 normalize, n
6300 Normalize biquad coefficients, by default is disabled. Enabling it
6301 will normalize magnitude response at DC to 0dB.
6302 transform, a
6304 Set transform type of IIR filter.
6305 di
6307 dii
6308 tdi
6309 tdii
6310 latt
6311 svf
6312 zdf
6313 precision, r
6314 Set precison of filtering.
6315 auto
6317 Pick automatic sample format depending on surround filters.
6318 s16 Always use signed 16-bit.
6320 s32 Always use signed 32-bit.
6322 f32 Always use float 32-bit.
6324 f64 Always use float 64-bit.
6326 block_size, b
6328 Set block size used for reverse IIR processing. If this value is
6329 set to high enough value (higher than impulse response length
6330 truncated when reaches near zero values) filtering will become
6331 linear phase otherwise if not big enough it will just produce nasty
6332 artifacts.
6333 Note that filter delay will be exactly this many samples when set
6335 to non-zero value.
6336 Commands
6338 This filter supports some options as commands.
6340 treble, highshelf
6342 Boost or cut treble (upper) frequencies of the audio using a two-pole
6343 shelving filter with a response similar to that of a standard hi-fi's
6344 tone-controls. This is also known as shelving equalisation (EQ).
6345 The filter accepts the following options:
6347 gain, g
6349 Give the gain at whichever is the lower of ~22 kHz and the Nyquist
6350 frequency. Its useful range is about -20 (for a large cut) to +20
6351 (for a large boost). Beware of clipping when using a positive gain.
6352 frequency, f
6354 Set the filter's central frequency and so can be used to extend or
6355 reduce the frequency range to be boosted or cut. The default value
6356 is 3000 Hz.
6357 width_type, t
6359 Set method to specify band-width of filter.
6360 h Hz
6362 q Q-Factor
6364 o octave
6366 s slope
6368 k kHz
6370 width, w
6372 Determine how steep is the filter's shelf transition.
6373 poles, p
6375 Set number of poles. Default is 2.
6376 mix, m
6378 How much to use filtered signal in output. Default is 1. Range is
6379 between 0 and 1.
6380 channels, c
6382 Specify which channels to filter, by default all available are
6383 filtered.
6384 normalize, n
6386 Normalize biquad coefficients, by default is disabled. Enabling it
6387 will normalize magnitude response at DC to 0dB.
6388 transform, a
6390 Set transform type of IIR filter.
6391 di
6393 dii
6394 tdi
6395 tdii
6396 latt
6397 svf
6398 zdf
6399 precision, r
6400 Set precison of filtering.
6401 auto
6403 Pick automatic sample format depending on surround filters.
6404 s16 Always use signed 16-bit.
6406 s32 Always use signed 32-bit.
6408 f32 Always use float 32-bit.
6410 f64 Always use float 64-bit.
6412 block_size, b
6414 Set block size used for reverse IIR processing. If this value is
6415 set to high enough value (higher than impulse response length
6416 truncated when reaches near zero values) filtering will become
6417 linear phase otherwise if not big enough it will just produce nasty
6418 artifacts.
6419 Note that filter delay will be exactly this many samples when set
6421 to non-zero value.
6422 Commands
6424 This filter supports the following commands:
6426 frequency, f
6428 Change treble frequency. Syntax for the command is : "frequency"
6429 width_type, t
6431 Change treble width_type. Syntax for the command is : "width_type"
6432 width, w
6434 Change treble width. Syntax for the command is : "width"
6435 gain, g
6437 Change treble gain. Syntax for the command is : "gain"
6438 mix, m
6440 Change treble mix. Syntax for the command is : "mix"
6441 tremolo
6443 Sinusoidal amplitude modulation.
6444 The filter accepts the following options:
6446 f Modulation frequency in Hertz. Modulation frequencies in the
6448 subharmonic range (20 Hz or lower) will result in a tremolo effect.
6449 This filter may also be used as a ring modulator by specifying a
6450 modulation frequency higher than 20 Hz. Range is 0.1 - 20000.0.
6451 Default value is 5.0 Hz.
6452 d Depth of modulation as a percentage. Range is 0.0 - 1.0. Default
6454 value is 0.5.
6455 vibrato
6457 Sinusoidal phase modulation.
6458 The filter accepts the following options:
6460 f Modulation frequency in Hertz. Range is 0.1 - 20000.0. Default
6462 value is 5.0 Hz.
6463 d Depth of modulation as a percentage. Range is 0.0 - 1.0. Default
6465 value is 0.5.
6466 virtualbass
6468 Apply audio Virtual Bass filter.
6469 This filter accepts stereo input and produce stereo with LFE (2.1)
6471 channels output. The newly produced LFE channel have enhanced virtual
6472 bass originally obtained from both stereo channels. This filter
6473 outputs front left and front right channels unchanged as available in
6474 stereo input.
6475 The filter accepts the following options:
6477 cutoff
6479 Set the virtual bass cutoff frequency. Default value is 250 Hz.
6480 Allowed range is from 100 to 500 Hz.
6481 strength
6483 Set the virtual bass strength. Allowed range is from 0.5 to 3.
6484 Default value is 3.
6485 volume
6487 Adjust the input audio volume.
6488 It accepts the following parameters:
6490 volume
6492 Set audio volume expression.
6493 Output values are clipped to the maximum value.
6495 The output audio volume is given by the relation:
6497 <output_volume> = <volume> * <input_volume>
6499 The default value for volume is "1.0".
6501 precision
6503 This parameter represents the mathematical precision.
6504 It determines which input sample formats will be allowed, which
6506 affects the precision of the volume scaling.
6507 fixed
6509 8-bit fixed-point; this limits input sample format to U8, S16,
6510 and S32.
6511 float
6513 32-bit floating-point; this limits input sample format to FLT.
6514 (default)
6515 double
6517 64-bit floating-point; this limits input sample format to DBL.
6518 replaygain
6520 Choose the behaviour on encountering ReplayGain side data in input
6521 frames.
6522 drop
6524 Remove ReplayGain side data, ignoring its contents (the
6525 default).
6526 ignore
6528 Ignore ReplayGain side data, but leave it in the frame.
6529 track
6531 Prefer the track gain, if present.
6532 album
6534 Prefer the album gain, if present.
6535 replaygain_preamp
6537 Pre-amplification gain in dB to apply to the selected replaygain
6538 gain.
6539 Default value for replaygain_preamp is 0.0.
6541 replaygain_noclip
6543 Prevent clipping by limiting the gain applied.
6544 Default value for replaygain_noclip is 1.
6546 eval
6548 Set when the volume expression is evaluated.
6549 It accepts the following values:
6551 once
6553 only evaluate expression once during the filter initialization,
6554 or when the volume command is sent
6555 frame
6557 evaluate expression for each incoming frame
6558 Default value is once.
6560 The volume expression can contain the following parameters.
6562 n frame number (starting at zero)
6564 nb_channels
6566 number of channels
6567 nb_consumed_samples
6569 number of samples consumed by the filter
6570 nb_samples
6572 number of samples in the current frame
6573 pos original frame position in the file
6575 pts frame PTS
6577 sample_rate
6579 sample rate
6580 startpts
6582 PTS at start of stream
6583 startt
6585 time at start of stream
6586 t frame time
6588 tb timestamp timebase
6590 volume
6592 last set volume value
6593 Note that when eval is set to once only the sample_rate and tb
6595 variables are available, all other variables will evaluate to NAN.
6596 Commands
6598 This filter supports the following commands:
6600 volume
6602 Modify the volume expression. The command accepts the same syntax
6603 of the corresponding option.
6604 If the specified expression is not valid, it is kept at its current
6606 value.
6607 Examples
6609 • Halve the input audio volume:
6611 volume=volume=0.5
6613 volume=volume=1/2
6614 volume=volume=-6.0206dB
6615 In all the above example the named key for volume can be omitted,
6617 for example like in:
6618 volume=0.5
6620 • Increase input audio power by 6 decibels using fixed-point
6622 precision:
6623 volume=volume=6dB:precision=fixed
6625 • Fade volume after time 10 with an annihilation period of 5 seconds:
6627 volume='if(lt(t,10),1,max(1-(t-10)/5,0))':eval=frame
6629 volumedetect
6631 Detect the volume of the input video.
6632 The filter has no parameters. It supports only 16-bit signed integer
6634 samples, so the input will be converted when needed. Statistics about
6635 the volume will be printed in the log when the input stream end is
6636 reached.
6637 In particular it will show the mean volume (root mean square), maximum
6639 volume (on a per-sample basis), and the beginning of a histogram of the
6640 registered volume values (from the maximum value to a cumulated 1/1000
6641 of the samples).
6642 All volumes are in decibels relative to the maximum PCM value.
6644 Examples
6646 Here is an excerpt of the output:
6648 [Parsed_volumedetect_0 0xa23120] mean_volume: -27 dB
6650 [Parsed_volumedetect_0 0xa23120] max_volume: -4 dB
6651 [Parsed_volumedetect_0 0xa23120] histogram_4db: 6
6652 [Parsed_volumedetect_0 0xa23120] histogram_5db: 62
6653 [Parsed_volumedetect_0 0xa23120] histogram_6db: 286
6654 [Parsed_volumedetect_0 0xa23120] histogram_7db: 1042
6655 [Parsed_volumedetect_0 0xa23120] histogram_8db: 2551
6656 [Parsed_volumedetect_0 0xa23120] histogram_9db: 4609
6657 [Parsed_volumedetect_0 0xa23120] histogram_10db: 8409
6658 It means that:
6660 • The mean square energy is approximately -27 dB, or 10^-2.7.
6662 • The largest sample is at -4 dB, or more precisely between -4 dB and
6664 -5 dB.
6665 • There are 6 samples at -4 dB, 62 at -5 dB, 286 at -6 dB, etc.
6667 In other words, raising the volume by +4 dB does not cause any
6669 clipping, raising it by +5 dB causes clipping for 6 samples, etc.
6670
6672 Below is a description of the currently available audio sources.
6673 abuffer
6675 Buffer audio frames, and make them available to the filter chain.
6676 This source is mainly intended for a programmatic use, in particular
6678 through the interface defined in libavfilter/buffersrc.h.
6679 It accepts the following parameters:
6681 time_base
6683 The timebase which will be used for timestamps of submitted frames.
6684 It must be either a floating-point number or in
6685 numerator/denominator form.
6686 sample_rate
6688 The sample rate of the incoming audio buffers.
6689 sample_fmt
6691 The sample format of the incoming audio buffers. Either a sample
6692 format name or its corresponding integer representation from the
6693 enum AVSampleFormat in libavutil/samplefmt.h
6694 channel_layout
6696 The channel layout of the incoming audio buffers. Either a channel
6697 layout name from channel_layout_map in libavutil/channel_layout.c
6698 or its corresponding integer representation from the AV_CH_LAYOUT_*
6699 macros in libavutil/channel_layout.h
6700 channels
6702 The number of channels of the incoming audio buffers. If both
6703 channels and channel_layout are specified, then they must be
6704 consistent.
6705 Examples
6707 abuffer=sample_rate=44100:sample_fmt=s16p:channel_layout=stereo
6709 will instruct the source to accept planar 16bit signed stereo at
6711 44100Hz. Since the sample format with name "s16p" corresponds to the
6712 number 6 and the "stereo" channel layout corresponds to the value 0x3,
6713 this is equivalent to:
6714 abuffer=sample_rate=44100:sample_fmt=6:channel_layout=0x3
6716 aevalsrc
6718 Generate an audio signal specified by an expression.
6719 This source accepts in input one or more expressions (one for each
6721 channel), which are evaluated and used to generate a corresponding
6722 audio signal.
6723 This source accepts the following options:
6725 exprs
6727 Set the '|'-separated expressions list for each separate channel.
6728 In case the channel_layout option is not specified, the selected
6729 channel layout depends on the number of provided expressions.
6730 Otherwise the last specified expression is applied to the remaining
6731 output channels.
6732 channel_layout, c
6734 Set the channel layout. The number of channels in the specified
6735 layout must be equal to the number of specified expressions.
6736 duration, d
6738 Set the minimum duration of the sourced audio. See the Time
6739 duration section in the ffmpeg-utils(1) manual for the accepted
6740 syntax. Note that the resulting duration may be greater than the
6741 specified duration, as the generated audio is always cut at the end
6742 of a complete frame.
6743 If not specified, or the expressed duration is negative, the audio
6745 is supposed to be generated forever.
6746 nb_samples, n
6748 Set the number of samples per channel per each output frame,
6749 default to 1024.
6750 sample_rate, s
6752 Specify the sample rate, default to 44100.
6753 Each expression in exprs can contain the following constants:
6755 n number of the evaluated sample, starting from 0
6757 t time of the evaluated sample expressed in seconds, starting from 0
6759 s sample rate
6761 Examples
6763 • Generate silence:
6765 aevalsrc=0
6767 • Generate a sin signal with frequency of 440 Hz, set sample rate to
6769 8000 Hz:
6770 aevalsrc="sin(440*2*PI*t):s=8000"
6772 • Generate a two channels signal, specify the channel layout (Front
6774 Center + Back Center) explicitly:
6775 aevalsrc="sin(420*2*PI*t)|cos(430*2*PI*t):c=FC|BC"
6777 • Generate white noise:
6779 aevalsrc="-2+random(0)"
6781 • Generate an amplitude modulated signal:
6783 aevalsrc="sin(10*2*PI*t)*sin(880*2*PI*t)"
6785 • Generate 2.5 Hz binaural beats on a 360 Hz carrier:
6787 aevalsrc="0.1*sin(2*PI*(360-2.5/2)*t) | 0.1*sin(2*PI*(360+2.5/2)*t)"
6789 afdelaysrc
6791 Generate a fractional delay FIR coefficients.
6792 The resulting stream can be used with afir filter for filtering the
6794 audio signal.
6795 The filter accepts the following options:
6797 delay, d
6799 Set the fractional delay. Default is 0.
6800 sample_rate, r
6802 Set the sample rate, default is 44100.
6803 nb_samples, n
6805 Set the number of samples per each frame. Default is 1024.
6806 taps, t
6808 Set the number of filter coefficents in output audio stream.
6809 Default value is 0.
6810 channel_layout, c
6812 Specifies the channel layout, and can be a string representing a
6813 channel layout. The default value of channel_layout is "stereo".
6814 afirsrc
6816 Generate a FIR coefficients using frequency sampling method.
6817 The resulting stream can be used with afir filter for filtering the
6819 audio signal.
6820 The filter accepts the following options:
6822 taps, t
6824 Set number of filter coefficents in output audio stream. Default
6825 value is 1025.
6826 frequency, f
6828 Set frequency points from where magnitude and phase are set. This
6829 must be in non decreasing order, and first element must be 0, while
6830 last element must be 1. Elements are separated by white spaces.
6831 magnitude, m
6833 Set magnitude value for every frequency point set by frequency.
6834 Number of values must be same as number of frequency points.
6835 Values are separated by white spaces.
6836 phase, p
6838 Set phase value for every frequency point set by frequency. Number
6839 of values must be same as number of frequency points. Values are
6840 separated by white spaces.
6841 sample_rate, r
6843 Set sample rate, default is 44100.
6844 nb_samples, n
6846 Set number of samples per each frame. Default is 1024.
6847 win_func, w
6849 Set window function. Default is blackman.
6850 anullsrc
6852 The null audio source, return unprocessed audio frames. It is mainly
6853 useful as a template and to be employed in analysis / debugging tools,
6854 or as the source for filters which ignore the input data (for example
6855 the sox synth filter).
6856 This source accepts the following options:
6858 channel_layout, cl
6860 Specifies the channel layout, and can be either an integer or a
6861 string representing a channel layout. The default value of
6862 channel_layout is "stereo".
6863 Check the channel_layout_map definition in
6865 libavutil/channel_layout.c for the mapping between strings and
6866 channel layout values.
6867 sample_rate, r
6869 Specifies the sample rate, and defaults to 44100.
6870 nb_samples, n
6872 Set the number of samples per requested frames.
6873 duration, d
6875 Set the duration of the sourced audio. See the Time duration
6876 section in the ffmpeg-utils(1) manual for the accepted syntax.
6877 If not specified, or the expressed duration is negative, the audio
6879 is supposed to be generated forever.
6880 Examples
6882 • Set the sample rate to 48000 Hz and the channel layout to
6884 AV_CH_LAYOUT_MONO.
6885 anullsrc=r=48000:cl=4
6887 • Do the same operation with a more obvious syntax:
6889 anullsrc=r=48000:cl=mono
6891 All the parameters need to be explicitly defined.
6893 flite
6895 Synthesize a voice utterance using the libflite library.
6896 To enable compilation of this filter you need to configure FFmpeg with
6898 "--enable-libflite".
6899 Note that versions of the flite library prior to 2.0 are not thread-
6901 safe.
6902 The filter accepts the following options:
6904 list_voices
6906 If set to 1, list the names of the available voices and exit
6907 immediately. Default value is 0.
6908 nb_samples, n
6910 Set the maximum number of samples per frame. Default value is 512.
6911 textfile
6913 Set the filename containing the text to speak.
6914 text
6916 Set the text to speak.
6917 voice, v
6919 Set the voice to use for the speech synthesis. Default value is
6920 "kal". See also the list_voices option.
6921 Examples
6923 • Read from file speech.txt, and synthesize the text using the
6925 standard flite voice:
6926 flite=textfile=speech.txt
6928 • Read the specified text selecting the "slt" voice:
6930 flite=text='So fare thee well, poor devil of a Sub-Sub, whose commentator I am':voice=slt
6932 • Input text to ffmpeg:
6934 ffmpeg -f lavfi -i flite=text='So fare thee well, poor devil of a Sub-Sub, whose commentator I am':voice=slt
6936 • Make ffplay speak the specified text, using "flite" and the "lavfi"
6938 device:
6939 ffplay -f lavfi flite=text='No more be grieved for which that thou hast done.'
6941 For more information about libflite, check:
6943 <http://www.festvox.org/flite/>
6944 anoisesrc
6946 Generate a noise audio signal.
6947 The filter accepts the following options:
6949 sample_rate, r
6951 Specify the sample rate. Default value is 48000 Hz.
6952 amplitude, a
6954 Specify the amplitude (0.0 - 1.0) of the generated audio stream.
6955 Default value is 1.0.
6956 duration, d
6958 Specify the duration of the generated audio stream. Not specifying
6959 this option results in noise with an infinite length.
6960 color, colour, c
6962 Specify the color of noise. Available noise colors are white, pink,
6963 brown, blue, violet and velvet. Default color is white.
6964 seed, s
6966 Specify a value used to seed the PRNG.
6967 nb_samples, n
6969 Set the number of samples per each output frame, default is 1024.
6970 Examples
6972 • Generate 60 seconds of pink noise, with a 44.1 kHz sampling rate
6974 and an amplitude of 0.5:
6975 anoisesrc=d=60:c=pink:r=44100:a=0.5
6977 hilbert
6979 Generate odd-tap Hilbert transform FIR coefficients.
6980 The resulting stream can be used with afir filter for phase-shifting
6982 the signal by 90 degrees.
6983 This is used in many matrix coding schemes and for analytic signal
6985 generation. The process is often written as a multiplication by i (or
6986 j), the imaginary unit.
6987 The filter accepts the following options:
6989 sample_rate, s
6991 Set sample rate, default is 44100.
6992 taps, t
6994 Set length of FIR filter, default is 22051.
6995 nb_samples, n
6997 Set number of samples per each frame.
6998 win_func, w
7000 Set window function to be used when generating FIR coefficients.
7001 sinc
7003 Generate a sinc kaiser-windowed low-pass, high-pass, band-pass, or
7004 band-reject FIR coefficients.
7005 The resulting stream can be used with afir filter for filtering the
7007 audio signal.
7008 The filter accepts the following options:
7010 sample_rate, r
7012 Set sample rate, default is 44100.
7013 nb_samples, n
7015 Set number of samples per each frame. Default is 1024.
7016 hp Set high-pass frequency. Default is 0.
7018 lp Set low-pass frequency. Default is 0. If high-pass frequency is
7020 lower than low-pass frequency and low-pass frequency is higher than
7021 0 then filter will create band-pass filter coefficients, otherwise
7022 band-reject filter coefficients.
7023 phase
7025 Set filter phase response. Default is 50. Allowed range is from 0
7026 to 100.
7027 beta
7029 Set Kaiser window beta.
7030 att Set stop-band attenuation. Default is 120dB, allowed range is from
7032 40 to 180 dB.
7033 round
7035 Enable rounding, by default is disabled.
7036 hptaps
7038 Set number of taps for high-pass filter.
7039 lptaps
7041 Set number of taps for low-pass filter.
7042 sine
7044 Generate an audio signal made of a sine wave with amplitude 1/8.
7045 The audio signal is bit-exact.
7047 The filter accepts the following options:
7049 frequency, f
7051 Set the carrier frequency. Default is 440 Hz.
7052 beep_factor, b
7054 Enable a periodic beep every second with frequency beep_factor
7055 times the carrier frequency. Default is 0, meaning the beep is
7056 disabled.
7057 sample_rate, r
7059 Specify the sample rate, default is 44100.
7060 duration, d
7062 Specify the duration of the generated audio stream.
7063 samples_per_frame
7065 Set the number of samples per output frame.
7066 The expression can contain the following constants:
7068 n The (sequential) number of the output audio frame, starting
7070 from 0.
7071 pts The PTS (Presentation TimeStamp) of the output audio frame,
7073 expressed in TB units.
7074 t The PTS of the output audio frame, expressed in seconds.
7076 TB The timebase of the output audio frames.
7078 Default is 1024.
7080 Examples
7082 • Generate a simple 440 Hz sine wave:
7084 sine
7086 • Generate a 220 Hz sine wave with a 880 Hz beep each second, for 5
7088 seconds:
7089 sine=220:4:d=5
7091 sine=f=220:b=4:d=5
7092 sine=frequency=220:beep_factor=4:duration=5
7093 • Generate a 1 kHz sine wave following "1602,1601,1602,1601,1602"
7095 NTSC pattern:
7096 sine=1000:samples_per_frame='st(0,mod(n,5)); 1602-not(not(eq(ld(0),1)+eq(ld(0),3)))'
7098
7100 Below is a description of the currently available audio sinks.
7101 abuffersink
7103 Buffer audio frames, and make them available to the end of filter
7104 chain.
7105 This sink is mainly intended for programmatic use, in particular
7107 through the interface defined in libavfilter/buffersink.h or the
7108 options system.
7109 It accepts a pointer to an AVABufferSinkContext structure, which
7111 defines the incoming buffers' formats, to be passed as the opaque
7112 parameter to "avfilter_init_filter" for initialization.
7113 anullsink
7115 Null audio sink; do absolutely nothing with the input audio. It is
7116 mainly useful as a template and for use in analysis / debugging tools.
7117
7119 When you configure your FFmpeg build, you can disable any of the
7120 existing filters using "--disable-filters". The configure output will
7121 show the video filters included in your build.
7122 Below is a description of the currently available video filters.
7124 addroi
7126 Mark a region of interest in a video frame.
7127 The frame data is passed through unchanged, but metadata is attached to
7129 the frame indicating regions of interest which can affect the behaviour
7130 of later encoding. Multiple regions can be marked by applying the
7131 filter multiple times.
7132 x Region distance in pixels from the left edge of the frame.
7134 y Region distance in pixels from the top edge of the frame.
7136 w Region width in pixels.
7138 h Region height in pixels.
7140 The parameters x, y, w and h are expressions, and may contain the
7142 following variables:
7143 iw Width of the input frame.
7145 ih Height of the input frame.
7147 qoffset
7149 Quantisation offset to apply within the region.
7150 This must be a real value in the range -1 to +1. A value of zero
7152 indicates no quality change. A negative value asks for better
7153 quality (less quantisation), while a positive value asks for worse
7154 quality (greater quantisation).
7155 The range is calibrated so that the extreme values indicate the
7157 largest possible offset - if the rest of the frame is encoded with
7158 the worst possible quality, an offset of -1 indicates that this
7159 region should be encoded with the best possible quality anyway.
7160 Intermediate values are then interpolated in some codec-dependent
7161 way.
7162 For example, in 10-bit H.264 the quantisation parameter varies
7164 between -12 and 51. A typical qoffset value of -1/10 therefore
7165 indicates that this region should be encoded with a QP around one-
7166 tenth of the full range better than the rest of the frame. So, if
7167 most of the frame were to be encoded with a QP of around 30, this
7168 region would get a QP of around 24 (an offset of approximately
7169 -1/10 * (51 - -12) = -6.3). An extreme value of -1 would indicate
7170 that this region should be encoded with the best possible quality
7171 regardless of the treatment of the rest of the frame - that is,
7172 should be encoded at a QP of -12.
7173 clear
7175 If set to true, remove any existing regions of interest marked on
7176 the frame before adding the new one.
7177 Examples
7179 • Mark the centre quarter of the frame as interesting.
7181 addroi=iw/4:ih/4:iw/2:ih/2:-1/10
7183 • Mark the 100-pixel-wide region on the left edge of the frame as
7185 very uninteresting (to be encoded at much lower quality than the
7186 rest of the frame).
7187 addroi=0:0:100:ih:+1/5
7189 alphaextract
7191 Extract the alpha component from the input as a grayscale video. This
7192 is especially useful with the alphamerge filter.
7193 alphamerge
7195 Add or replace the alpha component of the primary input with the
7196 grayscale value of a second input. This is intended for use with
7197 alphaextract to allow the transmission or storage of frame sequences
7198 that have alpha in a format that doesn't support an alpha channel.
7199 For example, to reconstruct full frames from a normal YUV-encoded video
7201 and a separate video created with alphaextract, you might use:
7202 movie=in_alpha.mkv [alpha]; [in][alpha] alphamerge [out]
7204 amplify
7206 Amplify differences between current pixel and pixels of adjacent frames
7207 in same pixel location.
7208 This filter accepts the following options:
7210 radius
7212 Set frame radius. Default is 2. Allowed range is from 1 to 63. For
7213 example radius of 3 will instruct filter to calculate average of 7
7214 frames.
7215 factor
7217 Set factor to amplify difference. Default is 2. Allowed range is
7218 from 0 to 65535.
7219 threshold
7221 Set threshold for difference amplification. Any difference greater
7222 or equal to this value will not alter source pixel. Default is 10.
7223 Allowed range is from 0 to 65535.
7224 tolerance
7226 Set tolerance for difference amplification. Any difference lower to
7227 this value will not alter source pixel. Default is 0. Allowed
7228 range is from 0 to 65535.
7229 low Set lower limit for changing source pixel. Default is 65535.
7231 Allowed range is from 0 to 65535. This option controls maximum
7232 possible value that will decrease source pixel value.
7233 high
7235 Set high limit for changing source pixel. Default is 65535. Allowed
7236 range is from 0 to 65535. This option controls maximum possible
7237 value that will increase source pixel value.
7238 planes
7240 Set which planes to filter. Default is all. Allowed range is from 0
7241 to 15.
7242 Commands
7244 This filter supports the following commands that corresponds to option
7246 of same name:
7247 factor
7249 threshold
7250 tolerance
7251 low
7252 high
7253 planes
7254 ass
7256 Same as the subtitles filter, except that it doesn't require libavcodec
7257 and libavformat to work. On the other hand, it is limited to ASS
7258 (Advanced Substation Alpha) subtitles files.
7259 This filter accepts the following option in addition to the common
7261 options from the subtitles filter:
7262 shaping
7264 Set the shaping engine
7265 Available values are:
7267 auto
7269 The default libass shaping engine, which is the best available.
7270 simple
7272 Fast, font-agnostic shaper that can do only substitutions
7273 complex
7275 Slower shaper using OpenType for substitutions and positioning
7276 The default is "auto".
7278 atadenoise
7280 Apply an Adaptive Temporal Averaging Denoiser to the video input.
7281 The filter accepts the following options:
7283 0a Set threshold A for 1st plane. Default is 0.02. Valid range is 0
7285 to 0.3.
7286 0b Set threshold B for 1st plane. Default is 0.04. Valid range is 0
7288 to 5.
7289 1a Set threshold A for 2nd plane. Default is 0.02. Valid range is 0
7291 to 0.3.
7292 1b Set threshold B for 2nd plane. Default is 0.04. Valid range is 0
7294 to 5.
7295 2a Set threshold A for 3rd plane. Default is 0.02. Valid range is 0
7297 to 0.3.
7298 2b Set threshold B for 3rd plane. Default is 0.04. Valid range is 0
7300 to 5.
7301 Threshold A is designed to react on abrupt changes in the input
7303 signal and threshold B is designed to react on continuous changes
7304 in the input signal.
7305 s Set number of frames filter will use for averaging. Default is 9.
7307 Must be odd number in range [5, 129].
7308 p Set what planes of frame filter will use for averaging. Default is
7310 all.
7311 a Set what variant of algorithm filter will use for averaging.
7313 Default is "p" parallel. Alternatively can be set to "s" serial.
7314 Parallel can be faster then serial, while other way around is never
7316 true. Parallel will abort early on first change being greater then
7317 thresholds, while serial will continue processing other side of
7318 frames if they are equal or below thresholds.
7319 0s
7321 1s
7322 2s Set sigma for 1st plane, 2nd plane or 3rd plane. Default is 32767.
7323 Valid range is from 0 to 32767. This options controls weight for
7324 each pixel in radius defined by size. Default value means every
7325 pixel have same weight. Setting this option to 0 effectively
7326 disables filtering.
7327 Commands
7329 This filter supports same commands as options except option "s". The
7331 command accepts the same syntax of the corresponding option.
7332 avgblur
7334 Apply average blur filter.
7335 The filter accepts the following options:
7337 sizeX
7339 Set horizontal radius size.
7340 planes
7342 Set which planes to filter. By default all planes are filtered.
7343 sizeY
7345 Set vertical radius size, if zero it will be same as "sizeX".
7346 Default is 0.
7347 Commands
7349 This filter supports same commands as options. The command accepts the
7351 same syntax of the corresponding option.
7352 If the specified expression is not valid, it is kept at its current
7354 value.
7355 backgroundkey
7357 Turns a static background into transparency.
7358 The filter accepts the following option:
7360 threshold
7362 Threshold for scene change detection.
7363 similarity
7365 Similarity percentage with the background.
7366 blend
7368 Set the blend amount for pixels that are not similar.
7369 Commands
7371 This filter supports the all above options as commands.
7373 bbox
7375 Compute the bounding box for the non-black pixels in the input frame
7376 luminance plane.
7377 This filter computes the bounding box containing all the pixels with a
7379 luminance value greater than the minimum allowed value. The parameters
7380 describing the bounding box are printed on the filter log.
7381 The filter accepts the following option:
7383 min_val
7385 Set the minimal luminance value. Default is 16.
7386 Commands
7388 This filter supports the all above options as commands.
7390 bilateral
7392 Apply bilateral filter, spatial smoothing while preserving edges.
7393 The filter accepts the following options:
7395 sigmaS
7397 Set sigma of gaussian function to calculate spatial weight.
7398 Allowed range is 0 to 512. Default is 0.1.
7399 sigmaR
7401 Set sigma of gaussian function to calculate range weight. Allowed
7402 range is 0 to 1. Default is 0.1.
7403 planes
7405 Set planes to filter. Default is first only.
7406 Commands
7408 This filter supports the all above options as commands.
7410 bilateral_cuda
7412 CUDA accelerated bilateral filter, an edge preserving filter. This
7413 filter is mathematically accurate thanks to the use of GPU
7414 acceleration. For best output quality, use one to one chroma
7415 subsampling, i.e. yuv444p format.
7416 The filter accepts the following options:
7418 sigmaS
7420 Set sigma of gaussian function to calculate spatial weight, also
7421 called sigma space. Allowed range is 0.1 to 512. Default is 0.1.
7422 sigmaR
7424 Set sigma of gaussian function to calculate color range weight,
7425 also called sigma color. Allowed range is 0.1 to 512. Default is
7426 0.1.
7427 window_size
7429 Set window size of the bilateral function to determine the number
7430 of neighbours to loop on. If the number entered is even, one will
7431 be added automatically. Allowed range is 1 to 255. Default is 1.
7432 Examples
7434 • Apply the bilateral filter on a video.
7436 ./ffmpeg -v verbose \
7438 -hwaccel cuda -hwaccel_output_format cuda -i input.mp4 \
7439 -init_hw_device cuda \
7440 -filter_complex \
7441 " \
7442 [0:v]scale_cuda=format=yuv444p[scaled_video];
7443 [scaled_video]bilateral_cuda=window_size=9:sigmaS=3.0:sigmaR=50.0" \
7444 -an -sn -c:v h264_nvenc -cq 20 out.mp4
7445 bitplanenoise
7447 Show and measure bit plane noise.
7448 The filter accepts the following options:
7450 bitplane
7452 Set which plane to analyze. Default is 1.
7453 filter
7455 Filter out noisy pixels from "bitplane" set above. Default is
7456 disabled.
7457 blackdetect
7459 Detect video intervals that are (almost) completely black. Can be
7460 useful to detect chapter transitions, commercials, or invalid
7461 recordings.
7462 The filter outputs its detection analysis to both the log as well as
7464 frame metadata. If a black segment of at least the specified minimum
7465 duration is found, a line with the start and end timestamps as well as
7466 duration is printed to the log with level "info". In addition, a log
7467 line with level "debug" is printed per frame showing the black amount
7468 detected for that frame.
7469 The filter also attaches metadata to the first frame of a black segment
7471 with key "lavfi.black_start" and to the first frame after the black
7472 segment ends with key "lavfi.black_end". The value is the frame's
7473 timestamp. This metadata is added regardless of the minimum duration
7474 specified.
7475 The filter accepts the following options:
7477 black_min_duration, d
7479 Set the minimum detected black duration expressed in seconds. It
7480 must be a non-negative floating point number.
7481 Default value is 2.0.
7483 picture_black_ratio_th, pic_th
7485 Set the threshold for considering a picture "black". Express the
7486 minimum value for the ratio:
7487 <nb_black_pixels> / <nb_pixels>
7489 for which a picture is considered black. Default value is 0.98.
7491 pixel_black_th, pix_th
7493 Set the threshold for considering a pixel "black".
7494 The threshold expresses the maximum pixel luminance value for which
7496 a pixel is considered "black". The provided value is scaled
7497 according to the following equation:
7498 <absolute_threshold> = <luminance_minimum_value> + <pixel_black_th> * <luminance_range_size>
7500 luminance_range_size and luminance_minimum_value depend on the
7502 input video format, the range is [0-255] for YUV full-range formats
7503 and [16-235] for YUV non full-range formats.
7504 Default value is 0.10.
7506 The following example sets the maximum pixel threshold to the minimum
7508 value, and detects only black intervals of 2 or more seconds:
7509 blackdetect=d=2:pix_th=0.00
7511 blackframe
7513 Detect frames that are (almost) completely black. Can be useful to
7514 detect chapter transitions or commercials. Output lines consist of the
7515 frame number of the detected frame, the percentage of blackness, the
7516 position in the file if known or -1 and the timestamp in seconds.
7517 In order to display the output lines, you need to set the loglevel at
7519 least to the AV_LOG_INFO value.
7520 This filter exports frame metadata "lavfi.blackframe.pblack". The
7522 value represents the percentage of pixels in the picture that are below
7523 the threshold value.
7524 It accepts the following parameters:
7526 amount
7528 The percentage of the pixels that have to be below the threshold;
7529 it defaults to 98.
7530 threshold, thresh
7532 The threshold below which a pixel value is considered black; it
7533 defaults to 32.
7534 blend
7536 Blend two video frames into each other.
7537 The "blend" filter takes two input streams and outputs one stream, the
7539 first input is the "top" layer and second input is "bottom" layer. By
7540 default, the output terminates when the longest input terminates.
7541 The "tblend" (time blend) filter takes two consecutive frames from one
7543 single stream, and outputs the result obtained by blending the new
7544 frame on top of the old frame.
7545 A description of the accepted options follows.
7547 c0_mode
7549 c1_mode
7550 c2_mode
7551 c3_mode
7552 all_mode
7553 Set blend mode for specific pixel component or all pixel components
7554 in case of all_mode. Default value is "normal".
7555 Available values for component modes are:
7557 addition
7559 and
7560 average
7561 bleach
7562 burn
7563 darken
7564 difference
7565 divide
7566 dodge
7567 exclusion
7568 extremity
7569 freeze
7570 geometric
7571 glow
7572 grainextract
7573 grainmerge
7574 hardlight
7575 hardmix
7576 hardoverlay
7577 harmonic
7578 heat
7579 interpolate
7580 lighten
7581 linearlight
7582 multiply
7583 multiply128
7584 negation
7585 normal
7586 or
7587 overlay
7588 phoenix
7589 pinlight
7590 reflect
7591 screen
7592 softdifference
7593 softlight
7594 stain
7595 subtract
7596 vividlight
7597 xor
7598 c0_opacity
7599 c1_opacity
7600 c2_opacity
7601 c3_opacity
7602 all_opacity
7603 Set blend opacity for specific pixel component or all pixel
7604 components in case of all_opacity. Only used in combination with
7605 pixel component blend modes.
7606 c0_expr
7608 c1_expr
7609 c2_expr
7610 c3_expr
7611 all_expr
7612 Set blend expression for specific pixel component or all pixel
7613 components in case of all_expr. Note that related mode options will
7614 be ignored if those are set.
7615 The expressions can use the following variables:
7617 N The sequential number of the filtered frame, starting from 0.
7619 X
7621 Y the coordinates of the current sample
7622 W
7624 H the width and height of currently filtered plane
7625 SW
7627 SH Width and height scale for the plane being filtered. It is the
7628 ratio between the dimensions of the current plane to the luma
7629 plane, e.g. for a "yuv420p" frame, the values are "1,1" for the
7630 luma plane and "0.5,0.5" for the chroma planes.
7631 T Time of the current frame, expressed in seconds.
7633 TOP, A
7635 Value of pixel component at current location for first video
7636 frame (top layer).
7637 BOTTOM, B
7639 Value of pixel component at current location for second video
7640 frame (bottom layer).
7641 The "blend" filter also supports the framesync options.
7643 Examples
7645 • Apply transition from bottom layer to top layer in first 10
7647 seconds:
7648 blend=all_expr='A*(if(gte(T,10),1,T/10))+B*(1-(if(gte(T,10),1,T/10)))'
7650 • Apply linear horizontal transition from top layer to bottom layer:
7652 blend=all_expr='A*(X/W)+B*(1-X/W)'
7654 • Apply 1x1 checkerboard effect:
7656 blend=all_expr='if(eq(mod(X,2),mod(Y,2)),A,B)'
7658 • Apply uncover left effect:
7660 blend=all_expr='if(gte(N*SW+X,W),A,B)'
7662 • Apply uncover down effect:
7664 blend=all_expr='if(gte(Y-N*SH,0),A,B)'
7666 • Apply uncover up-left effect:
7668 blend=all_expr='if(gte(T*SH*40+Y,H)*gte((T*40*SW+X)*W/H,W),A,B)'
7670 • Split diagonally video and shows top and bottom layer on each side:
7672 blend=all_expr='if(gt(X,Y*(W/H)),A,B)'
7674 • Display differences between the current and the previous frame:
7676 tblend=all_mode=grainextract
7678 Commands
7680 This filter supports same commands as options.
7682 blockdetect
7684 Determines blockiness of frames without altering the input frames.
7685 Based on Remco Muijs and Ihor Kirenko: "A no-reference blocking
7687 artifact measure for adaptive video processing." 2005 13th European
7688 signal processing conference.
7689 The filter accepts the following options:
7691 period_min
7693 period_max
7694 Set minimum and maximum values for determining pixel grids
7695 (periods). Default values are [3,24].
7696 planes
7698 Set planes to filter. Default is first only.
7699 Examples
7701 • Determine blockiness for the first plane and search for periods
7703 within [8,32]:
7704 blockdetect=period_min=8:period_max=32:planes=1
7706 blurdetect
7708 Determines blurriness of frames without altering the input frames.
7709 Based on Marziliano, Pina, et al. "A no-reference perceptual blur
7711 metric." Allows for a block-based abbreviation.
7712 The filter accepts the following options:
7714 low
7716 high
7717 Set low and high threshold values used by the Canny thresholding
7718 algorithm.
7719 The high threshold selects the "strong" edge pixels, which are then
7721 connected through 8-connectivity with the "weak" edge pixels
7722 selected by the low threshold.
7723 low and high threshold values must be chosen in the range [0,1],
7725 and low should be lesser or equal to high.
7726 Default value for low is "20/255", and default value for high is
7728 "50/255".
7729 radius
7731 Define the radius to search around an edge pixel for local maxima.
7732 block_pct
7734 Determine blurriness only for the most significant blocks, given in
7735 percentage.
7736 block_width
7738 Determine blurriness for blocks of width block_width. If set to any
7739 value smaller 1, no blocks are used and the whole image is
7740 processed as one no matter of block_height.
7741 block_height
7743 Determine blurriness for blocks of height block_height. If set to
7744 any value smaller 1, no blocks are used and the whole image is
7745 processed as one no matter of block_width.
7746 planes
7748 Set planes to filter. Default is first only.
7749 Examples
7751 • Determine blur for 80% of most significant 32x32 blocks:
7753 blurdetect=block_width=32:block_height=32:block_pct=80
7755 bm3d
7757 Denoise frames using Block-Matching 3D algorithm.
7758 The filter accepts the following options.
7760 sigma
7762 Set denoising strength. Default value is 1. Allowed range is from
7763 0 to 999.9. The denoising algorithm is very sensitive to sigma, so
7764 adjust it according to the source.
7765 block
7767 Set local patch size. This sets dimensions in 2D.
7768 bstep
7770 Set sliding step for processing blocks. Default value is 4.
7771 Allowed range is from 1 to 64. Smaller values allows processing
7772 more reference blocks and is slower.
7773 group
7775 Set maximal number of similar blocks for 3rd dimension. Default
7776 value is 1. When set to 1, no block matching is done. Larger
7777 values allows more blocks in single group. Allowed range is from 1
7778 to 256.
7779 range
7781 Set radius for search block matching. Default is 9. Allowed range
7782 is from 1 to INT32_MAX.
7783 mstep
7785 Set step between two search locations for block matching. Default
7786 is 1. Allowed range is from 1 to 64. Smaller is slower.
7787 thmse
7789 Set threshold of mean square error for block matching. Valid range
7790 is 0 to INT32_MAX.
7791 hdthr
7793 Set thresholding parameter for hard thresholding in 3D transformed
7794 domain. Larger values results in stronger hard-thresholding
7795 filtering in frequency domain.
7796 estim
7798 Set filtering estimation mode. Can be "basic" or "final". Default
7799 is "basic".
7800 ref If enabled, filter will use 2nd stream for block matching. Default
7802 is disabled for "basic" value of estim option, and always enabled
7803 if value of estim is "final".
7804 planes
7806 Set planes to filter. Default is all available except alpha.
7807 Examples
7809 • Basic filtering with bm3d:
7811 bm3d=sigma=3:block=4:bstep=2:group=1:estim=basic
7813 • Same as above, but filtering only luma:
7815 bm3d=sigma=3:block=4:bstep=2:group=1:estim=basic:planes=1
7817 • Same as above, but with both estimation modes:
7819 split[a][b],[a]bm3d=sigma=3:block=4:bstep=2:group=1:estim=basic[a],[b][a]bm3d=sigma=3:block=4:bstep=2:group=16:estim=final:ref=1
7821 • Same as above, but prefilter with nlmeans filter instead:
7823 split[a][b],[a]nlmeans=s=3:r=7:p=3[a],[b][a]bm3d=sigma=3:block=4:bstep=2:group=16:estim=final:ref=1
7825 boxblur
7827 Apply a boxblur algorithm to the input video.
7828 It accepts the following parameters:
7830 luma_radius, lr
7832 luma_power, lp
7833 chroma_radius, cr
7834 chroma_power, cp
7835 alpha_radius, ar
7836 alpha_power, ap
7837 A description of the accepted options follows.
7839 luma_radius, lr
7841 chroma_radius, cr
7842 alpha_radius, ar
7843 Set an expression for the box radius in pixels used for blurring
7844 the corresponding input plane.
7845 The radius value must be a non-negative number, and must not be
7847 greater than the value of the expression "min(w,h)/2" for the luma
7848 and alpha planes, and of "min(cw,ch)/2" for the chroma planes.
7849 Default value for luma_radius is "2". If not specified,
7851 chroma_radius and alpha_radius default to the corresponding value
7852 set for luma_radius.
7853 The expressions can contain the following constants:
7855 w
7857 h The input width and height in pixels.
7858 cw
7860 ch The input chroma image width and height in pixels.
7861 hsub
7863 vsub
7864 The horizontal and vertical chroma subsample values. For
7865 example, for the pixel format "yuv422p", hsub is 2 and vsub is
7866 1.
7867 luma_power, lp
7869 chroma_power, cp
7870 alpha_power, ap
7871 Specify how many times the boxblur filter is applied to the
7872 corresponding plane.
7873 Default value for luma_power is 2. If not specified, chroma_power
7875 and alpha_power default to the corresponding value set for
7876 luma_power.
7877 A value of 0 will disable the effect.
7879 Examples
7881 • Apply a boxblur filter with the luma, chroma, and alpha radii set
7883 to 2:
7884 boxblur=luma_radius=2:luma_power=1
7886 boxblur=2:1
7887 • Set the luma radius to 2, and alpha and chroma radius to 0:
7889 boxblur=2:1:cr=0:ar=0
7891 • Set the luma and chroma radii to a fraction of the video dimension:
7893 boxblur=luma_radius=min(h\,w)/10:luma_power=1:chroma_radius=min(cw\,ch)/10:chroma_power=1
7895 bwdif
7897 Deinterlace the input video ("bwdif" stands for "Bob Weaver
7898 Deinterlacing Filter").
7899 Motion adaptive deinterlacing based on yadif with the use of w3fdif and
7901 cubic interpolation algorithms. It accepts the following parameters:
7902 mode
7904 The interlacing mode to adopt. It accepts one of the following
7905 values:
7906 0, send_frame
7908 Output one frame for each frame.
7909 1, send_field
7911 Output one frame for each field.
7912 The default value is "send_field".
7914 parity
7916 The picture field parity assumed for the input interlaced video. It
7917 accepts one of the following values:
7918 0, tff
7920 Assume the top field is first.
7921 1, bff
7923 Assume the bottom field is first.
7924 -1, auto
7926 Enable automatic detection of field parity.
7927 The default value is "auto". If the interlacing is unknown or the
7929 decoder does not export this information, top field first will be
7930 assumed.
7931 deint
7933 Specify which frames to deinterlace. Accepts one of the following
7934 values:
7935 0, all
7937 Deinterlace all frames.
7938 1, interlaced
7940 Only deinterlace frames marked as interlaced.
7941 The default value is "all".
7943 cas
7945 Apply Contrast Adaptive Sharpen filter to video stream.
7946 The filter accepts the following options:
7948 strength
7950 Set the sharpening strength. Default value is 0.
7951 planes
7953 Set planes to filter. Default value is to filter all planes except
7954 alpha plane.
7955 Commands
7957 This filter supports same commands as options.
7959 chromahold
7961 Remove all color information for all colors except for certain one.
7962 The filter accepts the following options:
7964 color
7966 The color which will not be replaced with neutral chroma.
7967 similarity
7969 Similarity percentage with the above color. 0.01 matches only the
7970 exact key color, while 1.0 matches everything.
7971 blend
7973 Blend percentage. 0.0 makes pixels either fully gray, or not gray
7974 at all. Higher values result in more preserved color.
7975 yuv Signals that the color passed is already in YUV instead of RGB.
7977 Literal colors like "green" or "red" don't make sense with this
7979 enabled anymore. This can be used to pass exact YUV values as
7980 hexadecimal numbers.
7981 Commands
7983 This filter supports same commands as options. The command accepts the
7985 same syntax of the corresponding option.
7986 If the specified expression is not valid, it is kept at its current
7988 value.
7989 chromakey
7991 YUV colorspace color/chroma keying.
7992 The filter accepts the following options:
7994 color
7996 The color which will be replaced with transparency.
7997 similarity
7999 Similarity percentage with the key color.
8000 0.01 matches only the exact key color, while 1.0 matches
8002 everything.
8003 blend
8005 Blend percentage.
8006 0.0 makes pixels either fully transparent, or not transparent at
8008 all.
8009 Higher values result in semi-transparent pixels, with a higher
8011 transparency the more similar the pixels color is to the key color.
8012 yuv Signals that the color passed is already in YUV instead of RGB.
8014 Literal colors like "green" or "red" don't make sense with this
8016 enabled anymore. This can be used to pass exact YUV values as
8017 hexadecimal numbers.
8018 Commands
8020 This filter supports same commands as options. The command accepts the
8022 same syntax of the corresponding option.
8023 If the specified expression is not valid, it is kept at its current
8025 value.
8026 Examples
8028 • Make every green pixel in the input image transparent:
8030 ffmpeg -i input.png -vf chromakey=green out.png
8032 • Overlay a greenscreen-video on top of a static black background.
8034 ffmpeg -f lavfi -i color=c=black:s=1280x720 -i video.mp4 -shortest -filter_complex "[1:v]chromakey=0x70de77:0.1:0.2[ckout];[0:v][ckout]overlay[out]" -map "[out]" output.mkv
8036 chromakey_cuda
8038 CUDA accelerated YUV colorspace color/chroma keying.
8039 This filter works like normal chromakey filter but operates on CUDA
8041 frames. for more details and parameters see chromakey.
8042 Examples
8044 • Make all the green pixels in the input video transparent and use it
8046 as an overlay for another video:
8047 ./ffmpeg \
8049 -hwaccel cuda -hwaccel_output_format cuda -i input_green.mp4 \
8050 -hwaccel cuda -hwaccel_output_format cuda -i base_video.mp4 \
8051 -init_hw_device cuda \
8052 -filter_complex \
8053 " \
8054 [0:v]chromakey_cuda=0x25302D:0.1:0.12:1[overlay_video]; \
8055 [1:v]scale_cuda=format=yuv420p[base]; \
8056 [base][overlay_video]overlay_cuda" \
8057 -an -sn -c:v h264_nvenc -cq 20 output.mp4
8058 • Process two software sources, explicitly uploading the frames:
8060 ./ffmpeg -init_hw_device cuda=cuda -filter_hw_device cuda \
8062 -f lavfi -i color=size=800x600:color=white,format=yuv420p \
8063 -f lavfi -i yuvtestsrc=size=200x200,format=yuv420p \
8064 -filter_complex \
8065 " \
8066 [0]hwupload[under]; \
8067 [1]hwupload,chromakey_cuda=green:0.1:0.12[over]; \
8068 [under][over]overlay_cuda" \
8069 -c:v hevc_nvenc -cq 18 -preset slow output.mp4
8070 chromanr
8072 Reduce chrominance noise.
8073 The filter accepts the following options:
8075 thres
8077 Set threshold for averaging chrominance values. Sum of absolute
8078 difference of Y, U and V pixel components of current pixel and
8079 neighbour pixels lower than this threshold will be used in
8080 averaging. Luma component is left unchanged and is copied to
8081 output. Default value is 30. Allowed range is from 1 to 200.
8082 sizew
8084 Set horizontal radius of rectangle used for averaging. Allowed
8085 range is from 1 to 100. Default value is 5.
8086 sizeh
8088 Set vertical radius of rectangle used for averaging. Allowed range
8089 is from 1 to 100. Default value is 5.
8090 stepw
8092 Set horizontal step when averaging. Default value is 1. Allowed
8093 range is from 1 to 50. Mostly useful to speed-up filtering.
8094 steph
8096 Set vertical step when averaging. Default value is 1. Allowed
8097 range is from 1 to 50. Mostly useful to speed-up filtering.
8098 threy
8100 Set Y threshold for averaging chrominance values. Set finer
8101 control for max allowed difference between Y components of current
8102 pixel and neigbour pixels. Default value is 200. Allowed range is
8103 from 1 to 200.
8104 threu
8106 Set U threshold for averaging chrominance values. Set finer
8107 control for max allowed difference between U components of current
8108 pixel and neigbour pixels. Default value is 200. Allowed range is
8109 from 1 to 200.
8110 threv
8112 Set V threshold for averaging chrominance values. Set finer
8113 control for max allowed difference between V components of current
8114 pixel and neigbour pixels. Default value is 200. Allowed range is
8115 from 1 to 200.
8116 distance
8118 Set distance type used in calculations.
8119 manhattan
8121 Absolute difference.
8122 euclidean
8124 Difference squared.
8125 Default distance type is manhattan.
8127 Commands
8129 This filter supports same commands as options. The command accepts the
8131 same syntax of the corresponding option.
8132 chromashift
8134 Shift chroma pixels horizontally and/or vertically.
8135 The filter accepts the following options:
8137 cbh Set amount to shift chroma-blue horizontally.
8139 cbv Set amount to shift chroma-blue vertically.
8141 crh Set amount to shift chroma-red horizontally.
8143 crv Set amount to shift chroma-red vertically.
8145 edge
8147 Set edge mode, can be smear, default, or warp.
8148 Commands
8150 This filter supports the all above options as commands.
8152 ciescope
8154 Display CIE color diagram with pixels overlaid onto it.
8155 The filter accepts the following options:
8157 system
8159 Set color system.
8160 ntsc, 470m
8162 ebu, 470bg
8163 smpte
8164 240m
8165 apple
8166 widergb
8167 cie1931
8168 rec709, hdtv
8169 uhdtv, rec2020
8170 dcip3
8171 cie Set CIE system.
8172 xyy
8174 ucs
8175 luv
8176 gamuts
8177 Set what gamuts to draw.
8178 See "system" option for available values.
8180 size, s
8182 Set ciescope size, by default set to 512.
8183 intensity, i
8185 Set intensity used to map input pixel values to CIE diagram.
8186 contrast
8188 Set contrast used to draw tongue colors that are out of active
8189 color system gamut.
8190 corrgamma
8192 Correct gamma displayed on scope, by default enabled.
8193 showwhite
8195 Show white point on CIE diagram, by default disabled.
8196 gamma
8198 Set input gamma. Used only with XYZ input color space.
8199 fill
8201 Fill with CIE colors. By default is enabled.
8202 codecview
8204 Visualize information exported by some codecs.
8205 Some codecs can export information through frames using side-data or
8207 other means. For example, some MPEG based codecs export motion vectors
8208 through the export_mvs flag in the codec flags2 option.
8209 The filter accepts the following option:
8211 block
8213 Display block partition structure using the luma plane.
8214 mv Set motion vectors to visualize.
8216 Available flags for mv are:
8218 pf forward predicted MVs of P-frames
8220 bf forward predicted MVs of B-frames
8222 bb backward predicted MVs of B-frames
8224 qp Display quantization parameters using the chroma planes.
8226 mv_type, mvt
8228 Set motion vectors type to visualize. Includes MVs from all frames
8229 unless specified by frame_type option.
8230 Available flags for mv_type are:
8232 fp forward predicted MVs
8234 bp backward predicted MVs
8236 frame_type, ft
8238 Set frame type to visualize motion vectors of.
8239 Available flags for frame_type are:
8241 if intra-coded frames (I-frames)
8243 pf predicted frames (P-frames)
8245 bf bi-directionally predicted frames (B-frames)
8247 Examples
8249 • Visualize forward predicted MVs of all frames using ffplay:
8251 ffplay -flags2 +export_mvs input.mp4 -vf codecview=mv_type=fp
8253 • Visualize multi-directionals MVs of P and B-Frames using ffplay:
8255 ffplay -flags2 +export_mvs input.mp4 -vf codecview=mv=pf+bf+bb
8257 colorbalance
8259 Modify intensity of primary colors (red, green and blue) of input
8260 frames.
8261 The filter allows an input frame to be adjusted in the shadows,
8263 midtones or highlights regions for the red-cyan, green-magenta or blue-
8264 yellow balance.
8265 A positive adjustment value shifts the balance towards the primary
8267 color, a negative value towards the complementary color.
8268 The filter accepts the following options:
8270 rs
8272 gs
8273 bs Adjust red, green and blue shadows (darkest pixels).
8274 rm
8276 gm
8277 bm Adjust red, green and blue midtones (medium pixels).
8278 rh
8280 gh
8281 bh Adjust red, green and blue highlights (brightest pixels).
8282 Allowed ranges for options are "[-1.0, 1.0]". Defaults are 0.
8284 pl Preserve lightness when changing color balance. Default is
8286 disabled.
8287 Examples
8289 • Add red color cast to shadows:
8291 colorbalance=rs=.3
8293 Commands
8295 This filter supports the all above options as commands.
8297 colorcontrast
8299 Adjust color contrast between RGB components.
8300 The filter accepts the following options:
8302 rc Set the red-cyan contrast. Defaults is 0.0. Allowed range is from
8304 -1.0 to 1.0.
8305 gm Set the green-magenta contrast. Defaults is 0.0. Allowed range is
8307 from -1.0 to 1.0.
8308 by Set the blue-yellow contrast. Defaults is 0.0. Allowed range is
8310 from -1.0 to 1.0.
8311 rcw
8313 gmw
8314 byw Set the weight of each "rc", "gm", "by" option value. Default value
8315 is 0.0. Allowed range is from 0.0 to 1.0. If all weights are 0.0
8316 filtering is disabled.
8317 pl Set the amount of preserving lightness. Default value is 0.0.
8319 Allowed range is from 0.0 to 1.0.
8320 Commands
8322 This filter supports the all above options as commands.
8324 colorcorrect
8326 Adjust color white balance selectively for blacks and whites. This
8327 filter operates in YUV colorspace.
8328 The filter accepts the following options:
8330 rl Set the red shadow spot. Allowed range is from -1.0 to 1.0.
8332 Default value is 0.
8333 bl Set the blue shadow spot. Allowed range is from -1.0 to 1.0.
8335 Default value is 0.
8336 rh Set the red highlight spot. Allowed range is from -1.0 to 1.0.
8338 Default value is 0.
8339 bh Set the red highlight spot. Allowed range is from -1.0 to 1.0.
8341 Default value is 0.
8342 saturation
8344 Set the amount of saturation. Allowed range is from -3.0 to 3.0.
8345 Default value is 1.
8346 analyze
8348 If set to anything other than "manual" it will analyze every frame
8349 and use derived parameters for filtering output frame.
8350 Possible values are:
8352 manual
8354 average
8355 minmax
8356 median
8357 Default value is "manual".
8359 Commands
8361 This filter supports the all above options as commands.
8363 colorchannelmixer
8365 Adjust video input frames by re-mixing color channels.
8366 This filter modifies a color channel by adding the values associated to
8368 the other channels of the same pixels. For example if the value to
8369 modify is red, the output value will be:
8370 <red>=<red>*<rr> + <blue>*<rb> + <green>*<rg> + <alpha>*<ra>
8372 The filter accepts the following options:
8374 rr
8376 rg
8377 rb
8378 ra Adjust contribution of input red, green, blue and alpha channels
8379 for output red channel. Default is 1 for rr, and 0 for rg, rb and
8380 ra.
8381 gr
8383 gg
8384 gb
8385 ga Adjust contribution of input red, green, blue and alpha channels
8386 for output green channel. Default is 1 for gg, and 0 for gr, gb
8387 and ga.
8388 br
8390 bg
8391 bb
8392 ba Adjust contribution of input red, green, blue and alpha channels
8393 for output blue channel. Default is 1 for bb, and 0 for br, bg and
8394 ba.
8395 ar
8397 ag
8398 ab
8399 aa Adjust contribution of input red, green, blue and alpha channels
8400 for output alpha channel. Default is 1 for aa, and 0 for ar, ag
8401 and ab.
8402 Allowed ranges for options are "[-2.0, 2.0]".
8404 pc Set preserve color mode. The accepted values are:
8406 none
8408 Disable color preserving, this is default.
8409 lum Preserve luminance.
8411 max Preserve max value of RGB triplet.
8413 avg Preserve average value of RGB triplet.
8415 sum Preserve sum value of RGB triplet.
8417 nrm Preserve normalized value of RGB triplet.
8419 pwr Preserve power value of RGB triplet.
8421 pa Set the preserve color amount when changing colors. Allowed range
8423 is from "[0.0, 1.0]". Default is 0.0, thus disabled.
8424 Examples
8426 • Convert source to grayscale:
8428 colorchannelmixer=.3:.4:.3:0:.3:.4:.3:0:.3:.4:.3
8430 • Simulate sepia tones:
8432 colorchannelmixer=.393:.769:.189:0:.349:.686:.168:0:.272:.534:.131
8434 Commands
8436 This filter supports the all above options as commands.
8438 colorize
8440 Overlay a solid color on the video stream.
8441 The filter accepts the following options:
8443 hue Set the color hue. Allowed range is from 0 to 360. Default value
8445 is 0.
8446 saturation
8448 Set the color saturation. Allowed range is from 0 to 1. Default
8449 value is 0.5.
8450 lightness
8452 Set the color lightness. Allowed range is from 0 to 1. Default
8453 value is 0.5.
8454 mix Set the mix of source lightness. By default is set to 1.0. Allowed
8456 range is from 0.0 to 1.0.
8457 Commands
8459 This filter supports the all above options as commands.
8461 colorkey
8463 RGB colorspace color keying. This filter operates on 8-bit RGB format
8464 frames by setting the alpha component of each pixel which falls within
8465 the similarity radius of the key color to 0. The alpha value for pixels
8466 outside the similarity radius depends on the value of the blend option.
8467 The filter accepts the following options:
8469 color
8471 Set the color for which alpha will be set to 0 (full transparency).
8472 See "Color" section in the ffmpeg-utils manual. Default is
8473 "black".
8474 similarity
8476 Set the radius from the key color within which other colors also
8477 have full transparency. The computed distance is related to the
8478 unit fractional distance in 3D space between the RGB values of the
8479 key color and the pixel's color. Range is 0.01 to 1.0. 0.01 matches
8480 within a very small radius around the exact key color, while 1.0
8481 matches everything. Default is 0.01.
8482 blend
8484 Set how the alpha value for pixels that fall outside the similarity
8485 radius is computed. 0.0 makes pixels either fully transparent or
8486 fully opaque. Higher values result in semi-transparent pixels,
8487 with greater transparency the more similar the pixel color is to
8488 the key color. Range is 0.0 to 1.0. Default is 0.0.
8489 Examples
8491 • Make every green pixel in the input image transparent:
8493 ffmpeg -i input.png -vf colorkey=green out.png
8495 • Overlay a greenscreen-video on top of a static background image.
8497 ffmpeg -i background.png -i video.mp4 -filter_complex "[1:v]colorkey=0x3BBD1E:0.3:0.2[ckout];[0:v][ckout]overlay[out]" -map "[out]" output.flv
8499 Commands
8501 This filter supports same commands as options. The command accepts the
8503 same syntax of the corresponding option.
8504 If the specified expression is not valid, it is kept at its current
8506 value.
8507 colorhold
8509 Remove all color information for all RGB colors except for certain one.
8510 The filter accepts the following options:
8512 color
8514 The color which will not be replaced with neutral gray.
8515 similarity
8517 Similarity percentage with the above color. 0.01 matches only the
8518 exact key color, while 1.0 matches everything.
8519 blend
8521 Blend percentage. 0.0 makes pixels fully gray. Higher values
8522 result in more preserved color.
8523 Commands
8525 This filter supports same commands as options. The command accepts the
8527 same syntax of the corresponding option.
8528 If the specified expression is not valid, it is kept at its current
8530 value.
8531 colorlevels
8533 Adjust video input frames using levels.
8534 The filter accepts the following options:
8536 rimin
8538 gimin
8539 bimin
8540 aimin
8541 Adjust red, green, blue and alpha input black point. Allowed
8542 ranges for options are "[-1.0, 1.0]". Defaults are 0.
8543 rimax
8545 gimax
8546 bimax
8547 aimax
8548 Adjust red, green, blue and alpha input white point. Allowed
8549 ranges for options are "[-1.0, 1.0]". Defaults are 1.
8550 Input levels are used to lighten highlights (bright tones), darken
8552 shadows (dark tones), change the balance of bright and dark tones.
8553 romin
8555 gomin
8556 bomin
8557 aomin
8558 Adjust red, green, blue and alpha output black point. Allowed
8559 ranges for options are "[0, 1.0]". Defaults are 0.
8560 romax
8562 gomax
8563 bomax
8564 aomax
8565 Adjust red, green, blue and alpha output white point. Allowed
8566 ranges for options are "[0, 1.0]". Defaults are 1.
8567 Output levels allows manual selection of a constrained output level
8569 range.
8570 preserve
8572 Set preserve color mode. The accepted values are:
8573 none
8575 Disable color preserving, this is default.
8576 lum Preserve luminance.
8578 max Preserve max value of RGB triplet.
8580 avg Preserve average value of RGB triplet.
8582 sum Preserve sum value of RGB triplet.
8584 nrm Preserve normalized value of RGB triplet.
8586 pwr Preserve power value of RGB triplet.
8588 Examples
8590 • Make video output darker:
8592 colorlevels=rimin=0.058:gimin=0.058:bimin=0.058
8594 • Increase contrast:
8596 colorlevels=rimin=0.039:gimin=0.039:bimin=0.039:rimax=0.96:gimax=0.96:bimax=0.96
8598 • Make video output lighter:
8600 colorlevels=rimax=0.902:gimax=0.902:bimax=0.902
8602 • Increase brightness:
8604 colorlevels=romin=0.5:gomin=0.5:bomin=0.5
8606 Commands
8608 This filter supports the all above options as commands.
8610 colormap
8612 Apply custom color maps to video stream.
8613 This filter needs three input video streams. First stream is video
8615 stream that is going to be filtered out. Second and third video stream
8616 specify color patches for source color to target color mapping.
8617 The filter accepts the following options:
8619 patch_size
8621 Set the source and target video stream patch size in pixels.
8622 nb_patches
8624 Set the max number of used patches from source and target video
8625 stream. Default value is number of patches available in additional
8626 video streams. Max allowed number of patches is 64.
8627 type
8629 Set the adjustments used for target colors. Can be "relative" or
8630 "absolute". Defaults is "absolute".
8631 kernel
8633 Set the kernel used to measure color differences between mapped
8634 colors.
8635 The accepted values are:
8637 euclidean
8639 weuclidean
8640 Default is "euclidean".
8642 colormatrix
8644 Convert color matrix.
8645 The filter accepts the following options:
8647 src
8649 dst Specify the source and destination color matrix. Both values must
8650 be specified.
8651 The accepted values are:
8653 bt709
8655 BT.709
8656 fcc FCC
8658 bt601
8660 BT.601
8661 bt470
8663 BT.470
8664 bt470bg
8666 BT.470BG
8667 smpte170m
8669 SMPTE-170M
8670 smpte240m
8672 SMPTE-240M
8673 bt2020
8675 BT.2020
8676 For example to convert from BT.601 to SMPTE-240M, use the command:
8678 colormatrix=bt601:smpte240m
8680 colorspace
8682 Convert colorspace, transfer characteristics or color primaries. Input
8683 video needs to have an even size.
8684 The filter accepts the following options:
8686 all Specify all color properties at once.
8688 The accepted values are:
8690 bt470m
8692 BT.470M
8693 bt470bg
8695 BT.470BG
8696 bt601-6-525
8698 BT.601-6 525
8699 bt601-6-625
8701 BT.601-6 625
8702 bt709
8704 BT.709
8705 smpte170m
8707 SMPTE-170M
8708 smpte240m
8710 SMPTE-240M
8711 bt2020
8713 BT.2020
8714 space
8716 Specify output colorspace.
8717 The accepted values are:
8719 bt709
8721 BT.709
8722 fcc FCC
8724 bt470bg
8726 BT.470BG or BT.601-6 625
8727 smpte170m
8729 SMPTE-170M or BT.601-6 525
8730 smpte240m
8732 SMPTE-240M
8733 ycgco
8735 YCgCo
8736 bt2020ncl
8738 BT.2020 with non-constant luminance
8739 trc Specify output transfer characteristics.
8741 The accepted values are:
8743 bt709
8745 BT.709
8746 bt470m
8748 BT.470M
8749 bt470bg
8751 BT.470BG
8752 gamma22
8754 Constant gamma of 2.2
8755 gamma28
8757 Constant gamma of 2.8
8758 smpte170m
8760 SMPTE-170M, BT.601-6 625 or BT.601-6 525
8761 smpte240m
8763 SMPTE-240M
8764 srgb
8766 SRGB
8767 iec61966-2-1
8769 iec61966-2-1
8770 iec61966-2-4
8772 iec61966-2-4
8773 xvycc
8775 xvycc
8776 bt2020-10
8778 BT.2020 for 10-bits content
8779 bt2020-12
8781 BT.2020 for 12-bits content
8782 primaries
8784 Specify output color primaries.
8785 The accepted values are:
8787 bt709
8789 BT.709
8790 bt470m
8792 BT.470M
8793 bt470bg
8795 BT.470BG or BT.601-6 625
8796 smpte170m
8798 SMPTE-170M or BT.601-6 525
8799 smpte240m
8801 SMPTE-240M
8802 film
8804 film
8805 smpte431
8807 SMPTE-431
8808 smpte432
8810 SMPTE-432
8811 bt2020
8813 BT.2020
8814 jedec-p22
8816 JEDEC P22 phosphors
8817 range
8819 Specify output color range.
8820 The accepted values are:
8822 tv TV (restricted) range
8824 mpeg
8826 MPEG (restricted) range
8827 pc PC (full) range
8829 jpeg
8831 JPEG (full) range
8832 format
8834 Specify output color format.
8835 The accepted values are:
8837 yuv420p
8839 YUV 4:2:0 planar 8-bits
8840 yuv420p10
8842 YUV 4:2:0 planar 10-bits
8843 yuv420p12
8845 YUV 4:2:0 planar 12-bits
8846 yuv422p
8848 YUV 4:2:2 planar 8-bits
8849 yuv422p10
8851 YUV 4:2:2 planar 10-bits
8852 yuv422p12
8854 YUV 4:2:2 planar 12-bits
8855 yuv444p
8857 YUV 4:4:4 planar 8-bits
8858 yuv444p10
8860 YUV 4:4:4 planar 10-bits
8861 yuv444p12
8863 YUV 4:4:4 planar 12-bits
8864 fast
8866 Do a fast conversion, which skips gamma/primary correction. This
8867 will take significantly less CPU, but will be mathematically
8868 incorrect. To get output compatible with that produced by the
8869 colormatrix filter, use fast=1.
8870 dither
8872 Specify dithering mode.
8873 The accepted values are:
8875 none
8877 No dithering
8878 fsb Floyd-Steinberg dithering
8880 wpadapt
8882 Whitepoint adaptation mode.
8883 The accepted values are:
8885 bradford
8887 Bradford whitepoint adaptation
8888 vonkries
8890 von Kries whitepoint adaptation
8891 identity
8893 identity whitepoint adaptation (i.e. no whitepoint adaptation)
8894 iall
8896 Override all input properties at once. Same accepted values as all.
8897 ispace
8899 Override input colorspace. Same accepted values as space.
8900 iprimaries
8902 Override input color primaries. Same accepted values as primaries.
8903 itrc
8905 Override input transfer characteristics. Same accepted values as
8906 trc.
8907 irange
8909 Override input color range. Same accepted values as range.
8910 The filter converts the transfer characteristics, color space and color
8912 primaries to the specified user values. The output value, if not
8913 specified, is set to a default value based on the "all" property. If
8914 that property is also not specified, the filter will log an error. The
8915 output color range and format default to the same value as the input
8916 color range and format. The input transfer characteristics, color
8917 space, color primaries and color range should be set on the input data.
8918 If any of these are missing, the filter will log an error and no
8919 conversion will take place.
8920 For example to convert the input to SMPTE-240M, use the command:
8922 colorspace=smpte240m
8924 colorspace_cuda
8926 CUDA accelerated implementation of the colorspace filter.
8927 It is by no means feature complete compared to the software colorspace
8929 filter, and at the current time only supports color range conversion
8930 between jpeg/full and mpeg/limited range.
8931 The filter accepts the following options:
8933 range
8935 Specify output color range.
8936 The accepted values are:
8938 tv TV (restricted) range
8940 mpeg
8942 MPEG (restricted) range
8943 pc PC (full) range
8945 jpeg
8947 JPEG (full) range
8948 colortemperature
8950 Adjust color temperature in video to simulate variations in ambient
8951 color temperature.
8952 The filter accepts the following options:
8954 temperature
8956 Set the temperature in Kelvin. Allowed range is from 1000 to 40000.
8957 Default value is 6500 K.
8958 mix Set mixing with filtered output. Allowed range is from 0 to 1.
8960 Default value is 1.
8961 pl Set the amount of preserving lightness. Allowed range is from 0 to
8963 1. Default value is 0.
8964 Commands
8966 This filter supports same commands as options.
8968 convolution
8970 Apply convolution of 3x3, 5x5, 7x7 or horizontal/vertical up to 49
8971 elements.
8972 The filter accepts the following options:
8974 0m
8976 1m
8977 2m
8978 3m Set matrix for each plane. Matrix is sequence of 9, 25 or 49
8979 signed integers in square mode, and from 1 to 49 odd number of
8980 signed integers in row mode.
8981 0rdiv
8983 1rdiv
8984 2rdiv
8985 3rdiv
8986 Set multiplier for calculated value for each plane. If unset or 0,
8987 it will be sum of all matrix elements.
8988 0bias
8990 1bias
8991 2bias
8992 3bias
8993 Set bias for each plane. This value is added to the result of the
8994 multiplication. Useful for making the overall image brighter or
8995 darker. Default is 0.0.
8996 0mode
8998 1mode
8999 2mode
9000 3mode
9001 Set matrix mode for each plane. Can be square, row or column.
9002 Default is square.
9003 Commands
9005 This filter supports the all above options as commands.
9007 Examples
9009 • Apply sharpen:
9011 convolution="0 -1 0 -1 5 -1 0 -1 0:0 -1 0 -1 5 -1 0 -1 0:0 -1 0 -1 5 -1 0 -1 0:0 -1 0 -1 5 -1 0 -1 0"
9013 • Apply blur:
9015 convolution="1 1 1 1 1 1 1 1 1:1 1 1 1 1 1 1 1 1:1 1 1 1 1 1 1 1 1:1 1 1 1 1 1 1 1 1:1/9:1/9:1/9:1/9"
9017 • Apply edge enhance:
9019 convolution="0 0 0 -1 1 0 0 0 0:0 0 0 -1 1 0 0 0 0:0 0 0 -1 1 0 0 0 0:0 0 0 -1 1 0 0 0 0:5:1:1:1:0:128:128:128"
9021 • Apply edge detect:
9023 convolution="0 1 0 1 -4 1 0 1 0:0 1 0 1 -4 1 0 1 0:0 1 0 1 -4 1 0 1 0:0 1 0 1 -4 1 0 1 0:5:5:5:1:0:128:128:128"
9025 • Apply laplacian edge detector which includes diagonals:
9027 convolution="1 1 1 1 -8 1 1 1 1:1 1 1 1 -8 1 1 1 1:1 1 1 1 -8 1 1 1 1:1 1 1 1 -8 1 1 1 1:5:5:5:1:0:128:128:0"
9029 • Apply emboss:
9031 convolution="-2 -1 0 -1 1 1 0 1 2:-2 -1 0 -1 1 1 0 1 2:-2 -1 0 -1 1 1 0 1 2:-2 -1 0 -1 1 1 0 1 2"
9033 convolve
9035 Apply 2D convolution of video stream in frequency domain using second
9036 stream as impulse.
9037 The filter accepts the following options:
9039 planes
9041 Set which planes to process.
9042 impulse
9044 Set which impulse video frames will be processed, can be first or
9045 all. Default is all.
9046 The "convolve" filter also supports the framesync options.
9048 copy
9050 Copy the input video source unchanged to the output. This is mainly
9051 useful for testing purposes.
9052 coreimage
9054 Video filtering on GPU using Apple's CoreImage API on OSX.
9055 Hardware acceleration is based on an OpenGL context. Usually, this
9057 means it is processed by video hardware. However, software-based OpenGL
9058 implementations exist which means there is no guarantee for hardware
9059 processing. It depends on the respective OSX.
9060 There are many filters and image generators provided by Apple that come
9062 with a large variety of options. The filter has to be referenced by its
9063 name along with its options.
9064 The coreimage filter accepts the following options:
9066 list_filters
9068 List all available filters and generators along with all their
9069 respective options as well as possible minimum and maximum values
9070 along with the default values.
9071 list_filters=true
9073 filter
9075 Specify all filters by their respective name and options. Use
9076 list_filters to determine all valid filter names and options.
9077 Numerical options are specified by a float value and are
9078 automatically clamped to their respective value range. Vector and
9079 color options have to be specified by a list of space separated
9080 float values. Character escaping has to be done. A special option
9081 name "default" is available to use default options for a filter.
9082 It is required to specify either "default" or at least one of the
9084 filter options. All omitted options are used with their default
9085 values. The syntax of the filter string is as follows:
9086 filter=<NAME>@<OPTION>=<VALUE>[@<OPTION>=<VALUE>][@...][#<NAME>@<OPTION>=<VALUE>[@<OPTION>=<VALUE>][@...]][#...]
9088 output_rect
9090 Specify a rectangle where the output of the filter chain is copied
9091 into the input image. It is given by a list of space separated
9092 float values:
9093 output_rect=x\ y\ width\ height
9095 If not given, the output rectangle equals the dimensions of the
9097 input image. The output rectangle is automatically cropped at the
9098 borders of the input image. Negative values are valid for each
9099 component.
9100 output_rect=25\ 25\ 100\ 100
9102 Several filters can be chained for successive processing without GPU-
9104 HOST transfers allowing for fast processing of complex filter chains.
9105 Currently, only filters with zero (generators) or exactly one (filters)
9106 input image and one output image are supported. Also, transition
9107 filters are not yet usable as intended.
9108 Some filters generate output images with additional padding depending
9110 on the respective filter kernel. The padding is automatically removed
9111 to ensure the filter output has the same size as the input image.
9112 For image generators, the size of the output image is determined by the
9114 previous output image of the filter chain or the input image of the
9115 whole filterchain, respectively. The generators do not use the pixel
9116 information of this image to generate their output. However, the
9117 generated output is blended onto this image, resulting in partial or
9118 complete coverage of the output image.
9119 The coreimagesrc video source can be used for generating input images
9121 which are directly fed into the filter chain. By using it, providing
9122 input images by another video source or an input video is not required.
9123 Examples
9125 • List all filters available:
9127 coreimage=list_filters=true
9129 • Use the CIBoxBlur filter with default options to blur an image:
9131 coreimage=filter=CIBoxBlur@default
9133 • Use a filter chain with CISepiaTone at default values and
9135 CIVignetteEffect with its center at 100x100 and a radius of 50
9136 pixels:
9137 coreimage=filter=CIBoxBlur@default#CIVignetteEffect@inputCenter=100\ 100@inputRadius=50
9139 • Use nullsrc and CIQRCodeGenerator to create a QR code for the
9141 FFmpeg homepage, given as complete and escaped command-line for
9142 Apple's standard bash shell:
9143 ffmpeg -f lavfi -i nullsrc=s=100x100,coreimage=filter=CIQRCodeGenerator@inputMessage=https\\\\\://FFmpeg.org/@inputCorrectionLevel=H -frames:v 1 QRCode.png
9145 corr
9147 Obtain the correlation between two input videos.
9148 This filter takes two input videos.
9150 Both input videos must have the same resolution and pixel format for
9152 this filter to work correctly. Also it assumes that both inputs have
9153 the same number of frames, which are compared one by one.
9154 The obtained per component, average, min and max correlation is printed
9156 through the logging system.
9157 The filter stores the calculated correlation of each frame in frame
9159 metadata.
9160 This filter also supports the framesync options.
9162 In the below example the input file main.mpg being processed is
9164 compared with the reference file ref.mpg.
9165 ffmpeg -i main.mpg -i ref.mpg -lavfi corr -f null -
9167 cover_rect
9169 Cover a rectangular object
9170 It accepts the following options:
9172 cover
9174 Filepath of the optional cover image, needs to be in yuv420.
9175 mode
9177 Set covering mode.
9178 It accepts the following values:
9180 cover
9182 cover it by the supplied image
9183 blur
9185 cover it by interpolating the surrounding pixels
9186 Default value is blur.
9188 Examples
9190 • Cover a rectangular object by the supplied image of a given video
9192 using ffmpeg:
9193 ffmpeg -i file.ts -vf find_rect=newref.pgm,cover_rect=cover.jpg:mode=cover new.mkv
9195 crop
9197 Crop the input video to given dimensions.
9198 It accepts the following parameters:
9200 w, out_w
9202 The width of the output video. It defaults to "iw". This
9203 expression is evaluated only once during the filter configuration,
9204 or when the w or out_w command is sent.
9205 h, out_h
9207 The height of the output video. It defaults to "ih". This
9208 expression is evaluated only once during the filter configuration,
9209 or when the h or out_h command is sent.
9210 x The horizontal position, in the input video, of the left edge of
9212 the output video. It defaults to "(in_w-out_w)/2". This expression
9213 is evaluated per-frame.
9214 y The vertical position, in the input video, of the top edge of the
9216 output video. It defaults to "(in_h-out_h)/2". This expression is
9217 evaluated per-frame.
9218 keep_aspect
9220 If set to 1 will force the output display aspect ratio to be the
9221 same of the input, by changing the output sample aspect ratio. It
9222 defaults to 0.
9223 exact
9225 Enable exact cropping. If enabled, subsampled videos will be
9226 cropped at exact width/height/x/y as specified and will not be
9227 rounded to nearest smaller value. It defaults to 0.
9228 The out_w, out_h, x, y parameters are expressions containing the
9230 following constants:
9231 x
9233 y The computed values for x and y. They are evaluated for each new
9234 frame.
9235 in_w
9237 in_h
9238 The input width and height.
9239 iw
9241 ih These are the same as in_w and in_h.
9242 out_w
9244 out_h
9245 The output (cropped) width and height.
9246 ow
9248 oh These are the same as out_w and out_h.
9249 a same as iw / ih
9251 sar input sample aspect ratio
9253 dar input display aspect ratio, it is the same as (iw / ih) * sar
9255 hsub
9257 vsub
9258 horizontal and vertical chroma subsample values. For example for
9259 the pixel format "yuv422p" hsub is 2 and vsub is 1.
9260 n The number of the input frame, starting from 0.
9262 pos the position in the file of the input frame, NAN if unknown
9264 t The timestamp expressed in seconds. It's NAN if the input timestamp
9266 is unknown.
9267 The expression for out_w may depend on the value of out_h, and the
9269 expression for out_h may depend on out_w, but they cannot depend on x
9270 and y, as x and y are evaluated after out_w and out_h.
9271 The x and y parameters specify the expressions for the position of the
9273 top-left corner of the output (non-cropped) area. They are evaluated
9274 for each frame. If the evaluated value is not valid, it is approximated
9275 to the nearest valid value.
9276 The expression for x may depend on y, and the expression for y may
9278 depend on x.
9279 Examples
9281 • Crop area with size 100x100 at position (12,34).
9283 crop=100:100:12:34
9285 Using named options, the example above becomes:
9287 crop=w=100:h=100:x=12:y=34
9289 • Crop the central input area with size 100x100:
9291 crop=100:100
9293 • Crop the central input area with size 2/3 of the input video:
9295 crop=2/3*in_w:2/3*in_h
9297 • Crop the input video central square:
9299 crop=out_w=in_h
9301 crop=in_h
9302 • Delimit the rectangle with the top-left corner placed at position
9304 100:100 and the right-bottom corner corresponding to the right-
9305 bottom corner of the input image.
9306 crop=in_w-100:in_h-100:100:100
9308 • Crop 10 pixels from the left and right borders, and 20 pixels from
9310 the top and bottom borders
9311 crop=in_w-2*10:in_h-2*20
9313 • Keep only the bottom right quarter of the input image:
9315 crop=in_w/2:in_h/2:in_w/2:in_h/2
9317 • Crop height for getting Greek harmony:
9319 crop=in_w:1/PHI*in_w
9321 • Apply trembling effect:
9323 crop=in_w/2:in_h/2:(in_w-out_w)/2+((in_w-out_w)/2)*sin(n/10):(in_h-out_h)/2 +((in_h-out_h)/2)*sin(n/7)
9325 • Apply erratic camera effect depending on timestamp:
9327 crop=in_w/2:in_h/2:(in_w-out_w)/2+((in_w-out_w)/2)*sin(t*10):(in_h-out_h)/2 +((in_h-out_h)/2)*sin(t*13)"
9329 • Set x depending on the value of y:
9331 crop=in_w/2:in_h/2:y:10+10*sin(n/10)
9333 Commands
9335 This filter supports the following commands:
9337 w, out_w
9339 h, out_h
9340 x
9341 y Set width/height of the output video and the horizontal/vertical
9342 position in the input video. The command accepts the same syntax
9343 of the corresponding option.
9344 If the specified expression is not valid, it is kept at its current
9346 value.
9347 cropdetect
9349 Auto-detect the crop size.
9350 It calculates the necessary cropping parameters and prints the
9352 recommended parameters via the logging system. The detected dimensions
9353 correspond to the non-black or video area of the input video according
9354 to mode.
9355 It accepts the following parameters:
9357 mode
9359 Depending on mode crop detection is based on either the mere black
9360 value of surrounding pixels or a combination of motion vectors and
9361 edge pixels.
9362 black
9364 Detect black pixels surrounding the playing video. For fine
9365 control use option limit.
9366 mvedges
9368 Detect the playing video by the motion vectors inside the video
9369 and scanning for edge pixels typically forming the border of a
9370 playing video.
9371 limit
9373 Set higher black value threshold, which can be optionally specified
9374 from nothing (0) to everything (255 for 8-bit based formats). An
9375 intensity value greater to the set value is considered non-black.
9376 It defaults to 24. You can also specify a value between 0.0 and
9377 1.0 which will be scaled depending on the bitdepth of the pixel
9378 format.
9379 round
9381 The value which the width/height should be divisible by. It
9382 defaults to 16. The offset is automatically adjusted to center the
9383 video. Use 2 to get only even dimensions (needed for 4:2:2 video).
9384 16 is best when encoding to most video codecs.
9385 skip
9387 Set the number of initial frames for which evaluation is skipped.
9388 Default is 2. Range is 0 to INT_MAX.
9389 reset_count, reset
9391 Set the counter that determines after how many frames cropdetect
9392 will reset the previously detected largest video area and start
9393 over to detect the current optimal crop area. Default value is 0.
9394 This can be useful when channel logos distort the video area. 0
9396 indicates 'never reset', and returns the largest area encountered
9397 during playback.
9398 mv_threshold
9400 Set motion in pixel units as threshold for motion detection. It
9401 defaults to 8.
9402 low
9404 high
9405 Set low and high threshold values used by the Canny thresholding
9406 algorithm.
9407 The high threshold selects the "strong" edge pixels, which are then
9409 connected through 8-connectivity with the "weak" edge pixels
9410 selected by the low threshold.
9411 low and high threshold values must be chosen in the range [0,1],
9413 and low should be lesser or equal to high.
9414 Default value for low is "5/255", and default value for high is
9416 "15/255".
9417 Examples
9419 • Find video area surrounded by black borders:
9421 ffmpeg -i file.mp4 -vf cropdetect,metadata=mode=print -f null -
9423 • Find an embedded video area, generate motion vectors beforehand:
9425 ffmpeg -i file.mp4 -vf mestimate,cropdetect=mode=mvedges,metadata=mode=print -f null -
9427 • Find an embedded video area, use motion vectors from decoder:
9429 ffmpeg -flags2 +export_mvs -i file.mp4 -vf cropdetect=mode=mvedges,metadata=mode=print -f null -
9431 Commands
9433 This filter supports the following commands:
9435 limit
9437 The command accepts the same syntax of the corresponding option.
9438 If the specified expression is not valid, it is kept at its current
9439 value.
9440 cue
9442 Delay video filtering until a given wallclock timestamp. The filter
9443 first passes on preroll amount of frames, then it buffers at most
9444 buffer amount of frames and waits for the cue. After reaching the cue
9445 it forwards the buffered frames and also any subsequent frames coming
9446 in its input.
9447 The filter can be used synchronize the output of multiple ffmpeg
9449 processes for realtime output devices like decklink. By putting the
9450 delay in the filtering chain and pre-buffering frames the process can
9451 pass on data to output almost immediately after the target wallclock
9452 timestamp is reached.
9453 Perfect frame accuracy cannot be guaranteed, but the result is good
9455 enough for some use cases.
9456 cue The cue timestamp expressed in a UNIX timestamp in microseconds.
9458 Default is 0.
9459 preroll
9461 The duration of content to pass on as preroll expressed in seconds.
9462 Default is 0.
9463 buffer
9465 The maximum duration of content to buffer before waiting for the
9466 cue expressed in seconds. Default is 0.
9467 curves
9469 Apply color adjustments using curves.
9470 This filter is similar to the Adobe Photoshop and GIMP curves tools.
9472 Each component (red, green and blue) has its values defined by N key
9473 points tied from each other using a smooth curve. The x-axis represents
9474 the pixel values from the input frame, and the y-axis the new pixel
9475 values to be set for the output frame.
9476 By default, a component curve is defined by the two points (0;0) and
9478 (1;1). This creates a straight line where each original pixel value is
9479 "adjusted" to its own value, which means no change to the image.
9480 The filter allows you to redefine these two points and add some more. A
9482 new curve will be define to pass smoothly through all these new
9483 coordinates. The new defined points needs to be strictly increasing
9484 over the x-axis, and their x and y values must be in the [0;1]
9485 interval. The curve is formed by using a natural or monotonic cubic
9486 spline interpolation, depending on the interp option (default:
9487 "natural"). The "natural" spline produces a smoother curve in general
9488 while the monotonic ("pchip") spline guarantees the transitions between
9489 the specified points to be monotonic. If the computed curves happened
9490 to go outside the vector spaces, the values will be clipped
9491 accordingly.
9492 The filter accepts the following options:
9494 preset
9496 Select one of the available color presets. This option can be used
9497 in addition to the r, g, b parameters; in this case, the later
9498 options takes priority on the preset values. Available presets
9499 are:
9500 none
9502 color_negative
9503 cross_process
9504 darker
9505 increase_contrast
9506 lighter
9507 linear_contrast
9508 medium_contrast
9509 negative
9510 strong_contrast
9511 vintage
9512 Default is "none".
9514 master, m
9516 Set the master key points. These points will define a second pass
9517 mapping. It is sometimes called a "luminance" or "value" mapping.
9518 It can be used with r, g, b or all since it acts like a post-
9519 processing LUT.
9520 red, r
9522 Set the key points for the red component.
9523 green, g
9525 Set the key points for the green component.
9526 blue, b
9528 Set the key points for the blue component.
9529 all Set the key points for all components (not including master). Can
9531 be used in addition to the other key points component options. In
9532 this case, the unset component(s) will fallback on this all
9533 setting.
9534 psfile
9536 Specify a Photoshop curves file (".acv") to import the settings
9537 from.
9538 plot
9540 Save Gnuplot script of the curves in specified file.
9541 interp
9543 Specify the kind of interpolation. Available algorithms are:
9544 natural
9546 Natural cubic spline using a piece-wise cubic polynomial that
9547 is twice continuously differentiable.
9548 pchip
9550 Monotonic cubic spline using a piecewise cubic Hermite
9551 interpolating polynomial (PCHIP).
9552 To avoid some filtergraph syntax conflicts, each key points list need
9554 to be defined using the following syntax: "x0/y0 x1/y1 x2/y2 ...".
9555 Commands
9557 This filter supports same commands as options.
9559 Examples
9561 • Increase slightly the middle level of blue:
9563 curves=blue='0/0 0.5/0.58 1/1'
9565 • Vintage effect:
9567 curves=r='0/0.11 .42/.51 1/0.95':g='0/0 0.50/0.48 1/1':b='0/0.22 .49/.44 1/0.8'
9569 Here we obtain the following coordinates for each components:
9571 red "(0;0.11) (0.42;0.51) (1;0.95)"
9573 green
9575 "(0;0) (0.50;0.48) (1;1)"
9576 blue
9578 "(0;0.22) (0.49;0.44) (1;0.80)"
9579 • The previous example can also be achieved with the associated
9581 built-in preset:
9582 curves=preset=vintage
9584 • Or simply:
9586 curves=vintage
9588 • Use a Photoshop preset and redefine the points of the green
9590 component:
9591 curves=psfile='MyCurvesPresets/purple.acv':green='0/0 0.45/0.53 1/1'
9593 • Check out the curves of the "cross_process" profile using ffmpeg
9595 and gnuplot:
9596 ffmpeg -f lavfi -i color -vf curves=cross_process:plot=/tmp/curves.plt -frames:v 1 -f null -
9598 gnuplot -p /tmp/curves.plt
9599 datascope
9601 Video data analysis filter.
9602 This filter shows hexadecimal pixel values of part of video.
9604 The filter accepts the following options:
9606 size, s
9608 Set output video size.
9609 x Set x offset from where to pick pixels.
9611 y Set y offset from where to pick pixels.
9613 mode
9615 Set scope mode, can be one of the following:
9616 mono
9618 Draw hexadecimal pixel values with white color on black
9619 background.
9620 color
9622 Draw hexadecimal pixel values with input video pixel color on
9623 black background.
9624 color2
9626 Draw hexadecimal pixel values on color background picked from
9627 input video, the text color is picked in such way so its always
9628 visible.
9629 axis
9631 Draw rows and columns numbers on left and top of video.
9632 opacity
9634 Set background opacity.
9635 format
9637 Set display number format. Can be "hex", or "dec". Default is
9638 "hex".
9639 components
9641 Set pixel components to display. By default all pixel components
9642 are displayed.
9643 Commands
9645 This filter supports same commands as options excluding "size" option.
9647 dblur
9649 Apply Directional blur filter.
9650 The filter accepts the following options:
9652 angle
9654 Set angle of directional blur. Default is 45.
9655 radius
9657 Set radius of directional blur. Default is 5.
9658 planes
9660 Set which planes to filter. By default all planes are filtered.
9661 Commands
9663 This filter supports same commands as options. The command accepts the
9665 same syntax of the corresponding option.
9666 If the specified expression is not valid, it is kept at its current
9668 value.
9669 dctdnoiz
9671 Denoise frames using 2D DCT (frequency domain filtering).
9672 This filter is not designed for real time.
9674 The filter accepts the following options:
9676 sigma, s
9678 Set the noise sigma constant.
9679 This sigma defines a hard threshold of "3 * sigma"; every DCT
9681 coefficient (absolute value) below this threshold with be dropped.
9682 If you need a more advanced filtering, see expr.
9684 Default is 0.
9686 overlap
9688 Set number overlapping pixels for each block. Since the filter can
9689 be slow, you may want to reduce this value, at the cost of a less
9690 effective filter and the risk of various artefacts.
9691 If the overlapping value doesn't permit processing the whole input
9693 width or height, a warning will be displayed and according borders
9694 won't be denoised.
9695 Default value is blocksize-1, which is the best possible setting.
9697 expr, e
9699 Set the coefficient factor expression.
9700 For each coefficient of a DCT block, this expression will be
9702 evaluated as a multiplier value for the coefficient.
9703 If this is option is set, the sigma option will be ignored.
9705 The absolute value of the coefficient can be accessed through the c
9707 variable.
9708 n Set the blocksize using the number of bits. "1<<n" defines the
9710 blocksize, which is the width and height of the processed blocks.
9711 The default value is 3 (8x8) and can be raised to 4 for a blocksize
9713 of 16x16. Note that changing this setting has huge consequences on
9714 the speed processing. Also, a larger block size does not
9715 necessarily means a better de-noising.
9716 Examples
9718 Apply a denoise with a sigma of 4.5:
9720 dctdnoiz=4.5
9722 The same operation can be achieved using the expression system:
9724 dctdnoiz=e='gte(c, 4.5*3)'
9726 Violent denoise using a block size of "16x16":
9728 dctdnoiz=15:n=4
9730 deband
9732 Remove banding artifacts from input video. It works by replacing
9733 banded pixels with average value of referenced pixels.
9734 The filter accepts the following options:
9736 1thr
9738 2thr
9739 3thr
9740 4thr
9741 Set banding detection threshold for each plane. Default is 0.02.
9742 Valid range is 0.00003 to 0.5. If difference between current pixel
9743 and reference pixel is less than threshold, it will be considered
9744 as banded.
9745 range, r
9747 Banding detection range in pixels. Default is 16. If positive,
9748 random number in range 0 to set value will be used. If negative,
9749 exact absolute value will be used. The range defines square of
9750 four pixels around current pixel.
9751 direction, d
9753 Set direction in radians from which four pixel will be compared. If
9754 positive, random direction from 0 to set direction will be picked.
9755 If negative, exact of absolute value will be picked. For example
9756 direction 0, -PI or -2*PI radians will pick only pixels on same row
9757 and -PI/2 will pick only pixels on same column.
9758 blur, b
9760 If enabled, current pixel is compared with average value of all
9761 four surrounding pixels. The default is enabled. If disabled
9762 current pixel is compared with all four surrounding pixels. The
9763 pixel is considered banded if only all four differences with
9764 surrounding pixels are less than threshold.
9765 coupling, c
9767 If enabled, current pixel is changed if and only if all pixel
9768 components are banded, e.g. banding detection threshold is
9769 triggered for all color components. The default is disabled.
9770 Commands
9772 This filter supports the all above options as commands.
9774 deblock
9776 Remove blocking artifacts from input video.
9777 The filter accepts the following options:
9779 filter
9781 Set filter type, can be weak or strong. Default is strong. This
9782 controls what kind of deblocking is applied.
9783 block
9785 Set size of block, allowed range is from 4 to 512. Default is 8.
9786 alpha
9788 beta
9789 gamma
9790 delta
9791 Set blocking detection thresholds. Allowed range is 0 to 1.
9792 Defaults are: 0.098 for alpha and 0.05 for the rest. Using higher
9793 threshold gives more deblocking strength. Setting alpha controls
9794 threshold detection at exact edge of block. Remaining options
9795 controls threshold detection near the edge. Each one for
9796 below/above or left/right. Setting any of those to 0 disables
9797 deblocking.
9798 planes
9800 Set planes to filter. Default is to filter all available planes.
9801 Examples
9803 • Deblock using weak filter and block size of 4 pixels.
9805 deblock=filter=weak:block=4
9807 • Deblock using strong filter, block size of 4 pixels and custom
9809 thresholds for deblocking more edges.
9810 deblock=filter=strong:block=4:alpha=0.12:beta=0.07:gamma=0.06:delta=0.05
9812 • Similar as above, but filter only first plane.
9814 deblock=filter=strong:block=4:alpha=0.12:beta=0.07:gamma=0.06:delta=0.05:planes=1
9816 • Similar as above, but filter only second and third plane.
9818 deblock=filter=strong:block=4:alpha=0.12:beta=0.07:gamma=0.06:delta=0.05:planes=6
9820 Commands
9822 This filter supports the all above options as commands.
9824 decimate
9826 Drop duplicated frames at regular intervals.
9827 The filter accepts the following options:
9829 cycle
9831 Set the number of frames from which one will be dropped. Setting
9832 this to N means one frame in every batch of N frames will be
9833 dropped. Default is 5.
9834 dupthresh
9836 Set the threshold for duplicate detection. If the difference metric
9837 for a frame is less than or equal to this value, then it is
9838 declared as duplicate. Default is 1.1
9839 scthresh
9841 Set scene change threshold. Default is 15.
9842 blockx
9844 blocky
9845 Set the size of the x and y-axis blocks used during metric
9846 calculations. Larger blocks give better noise suppression, but
9847 also give worse detection of small movements. Must be a power of
9848 two. Default is 32.
9849 ppsrc
9851 Mark main input as a pre-processed input and activate clean source
9852 input stream. This allows the input to be pre-processed with
9853 various filters to help the metrics calculation while keeping the
9854 frame selection lossless. When set to 1, the first stream is for
9855 the pre-processed input, and the second stream is the clean source
9856 from where the kept frames are chosen. Default is 0.
9857 chroma
9859 Set whether or not chroma is considered in the metric calculations.
9860 Default is 1.
9861 mixed
9863 Set whether or not the input only partially contains content to be
9864 decimated. Default is "false". If enabled video output stream
9865 will be in variable frame rate.
9866 deconvolve
9868 Apply 2D deconvolution of video stream in frequency domain using second
9869 stream as impulse.
9870 The filter accepts the following options:
9872 planes
9874 Set which planes to process.
9875 impulse
9877 Set which impulse video frames will be processed, can be first or
9878 all. Default is all.
9879 noise
9881 Set noise when doing divisions. Default is 0.0000001. Useful when
9882 width and height are not same and not power of 2 or if stream prior
9883 to convolving had noise.
9884 The "deconvolve" filter also supports the framesync options.
9886 dedot
9888 Reduce cross-luminance (dot-crawl) and cross-color (rainbows) from
9889 video.
9890 It accepts the following options:
9892 m Set mode of operation. Can be combination of dotcrawl for cross-
9894 luminance reduction and/or rainbows for cross-color reduction.
9895 lt Set spatial luma threshold. Lower values increases reduction of
9897 cross-luminance.
9898 tl Set tolerance for temporal luma. Higher values increases reduction
9900 of cross-luminance.
9901 tc Set tolerance for chroma temporal variation. Higher values
9903 increases reduction of cross-color.
9904 ct Set temporal chroma threshold. Lower values increases reduction of
9906 cross-color.
9907 deflate
9909 Apply deflate effect to the video.
9910 This filter replaces the pixel by the local(3x3) average by taking into
9912 account only values lower than the pixel.
9913 It accepts the following options:
9915 threshold0
9917 threshold1
9918 threshold2
9919 threshold3
9920 Limit the maximum change for each plane, default is 65535. If 0,
9921 plane will remain unchanged.
9922 Commands
9924 This filter supports the all above options as commands.
9926 deflicker
9928 Remove temporal frame luminance variations.
9929 It accepts the following options:
9931 size, s
9933 Set moving-average filter size in frames. Default is 5. Allowed
9934 range is 2 - 129.
9935 mode, m
9937 Set averaging mode to smooth temporal luminance variations.
9938 Available values are:
9940 am Arithmetic mean
9942 gm Geometric mean
9944 hm Harmonic mean
9946 qm Quadratic mean
9948 cm Cubic mean
9950 pm Power mean
9952 median
9954 Median
9955 bypass
9957 Do not actually modify frame. Useful when one only wants metadata.
9958 dejudder
9960 Remove judder produced by partially interlaced telecined content.
9961 Judder can be introduced, for instance, by pullup filter. If the
9963 original source was partially telecined content then the output of
9964 "pullup,dejudder" will have a variable frame rate. May change the
9965 recorded frame rate of the container. Aside from that change, this
9966 filter will not affect constant frame rate video.
9967 The option available in this filter is:
9969 cycle
9971 Specify the length of the window over which the judder repeats.
9972 Accepts any integer greater than 1. Useful values are:
9974 4 If the original was telecined from 24 to 30 fps (Film to NTSC).
9976 5 If the original was telecined from 25 to 30 fps (PAL to NTSC).
9978 20 If a mixture of the two.
9980 The default is 4.
9982 delogo
9984 Suppress a TV station logo by a simple interpolation of the surrounding
9985 pixels. Just set a rectangle covering the logo and watch it disappear
9986 (and sometimes something even uglier appear - your mileage may vary).
9987 It accepts the following parameters:
9989 x
9991 y Specify the top left corner coordinates of the logo. They must be
9992 specified.
9993 w
9995 h Specify the width and height of the logo to clear. They must be
9996 specified.
9997 show
9999 When set to 1, a green rectangle is drawn on the screen to simplify
10000 finding the right x, y, w, and h parameters. The default value is
10001 0.
10002 The rectangle is drawn on the outermost pixels which will be
10004 (partly) replaced with interpolated values. The values of the next
10005 pixels immediately outside this rectangle in each direction will be
10006 used to compute the interpolated pixel values inside the rectangle.
10007 Examples
10009 • Set a rectangle covering the area with top left corner coordinates
10011 0,0 and size 100x77:
10012 delogo=x=0:y=0:w=100:h=77
10014 derain
10016 Remove the rain in the input image/video by applying the derain methods
10017 based on convolutional neural networks. Supported models:
10018 • Recurrent Squeeze-and-Excitation Context Aggregation Net (RESCAN).
10020 See
10021 <http://openaccess.thecvf.com/content_ECCV_2018/papers/Xia_Li_Recurrent_Squeeze-and-Excitation_Context_ECCV_2018_paper.pdf>.
10022 Training as well as model generation scripts are provided in the
10024 repository at <https://github.com/XueweiMeng/derain_filter.git>.
10025 Native model files (.model) can be generated from TensorFlow model
10027 files (.pb) by using tools/python/convert.py
10028 The filter accepts the following options:
10030 filter_type
10032 Specify which filter to use. This option accepts the following
10033 values:
10034 derain
10036 Derain filter. To conduct derain filter, you need to use a
10037 derain model.
10038 dehaze
10040 Dehaze filter. To conduct dehaze filter, you need to use a
10041 dehaze model.
10042 Default value is derain.
10044 dnn_backend
10046 Specify which DNN backend to use for model loading and execution.
10047 This option accepts the following values:
10048 native
10050 Native implementation of DNN loading and execution.
10051 tensorflow
10053 TensorFlow backend. To enable this backend you need to install
10054 the TensorFlow for C library (see
10055 <https://www.tensorflow.org/install/lang_c>) and configure
10056 FFmpeg with "--enable-libtensorflow"
10057 Default value is native.
10059 model
10061 Set path to model file specifying network architecture and its
10062 parameters. Note that different backends use different file
10063 formats. TensorFlow and native backend can load files for only its
10064 format.
10065 To get full functionality (such as async execution), please use the
10067 dnn_processing filter.
10068 deshake
10070 Attempt to fix small changes in horizontal and/or vertical shift. This
10071 filter helps remove camera shake from hand-holding a camera, bumping a
10072 tripod, moving on a vehicle, etc.
10073 The filter accepts the following options:
10075 x
10077 y
10078 w
10079 h Specify a rectangular area where to limit the search for motion
10080 vectors. If desired the search for motion vectors can be limited
10081 to a rectangular area of the frame defined by its top left corner,
10082 width and height. These parameters have the same meaning as the
10083 drawbox filter which can be used to visualise the position of the
10084 bounding box.
10085 This is useful when simultaneous movement of subjects within the
10087 frame might be confused for camera motion by the motion vector
10088 search.
10089 If any or all of x, y, w and h are set to -1 then the full frame is
10091 used. This allows later options to be set without specifying the
10092 bounding box for the motion vector search.
10093 Default - search the whole frame.
10095 rx
10097 ry Specify the maximum extent of movement in x and y directions in the
10098 range 0-64 pixels. Default 16.
10099 edge
10101 Specify how to generate pixels to fill blanks at the edge of the
10102 frame. Available values are:
10103 blank, 0
10105 Fill zeroes at blank locations
10106 original, 1
10108 Original image at blank locations
10109 clamp, 2
10111 Extruded edge value at blank locations
10112 mirror, 3
10114 Mirrored edge at blank locations
10115 Default value is mirror.
10117 blocksize
10119 Specify the blocksize to use for motion search. Range 4-128 pixels,
10120 default 8.
10121 contrast
10123 Specify the contrast threshold for blocks. Only blocks with more
10124 than the specified contrast (difference between darkest and
10125 lightest pixels) will be considered. Range 1-255, default 125.
10126 search
10128 Specify the search strategy. Available values are:
10129 exhaustive, 0
10131 Set exhaustive search
10132 less, 1
10134 Set less exhaustive search.
10135 Default value is exhaustive.
10137 filename
10139 If set then a detailed log of the motion search is written to the
10140 specified file.
10141 despill
10143 Remove unwanted contamination of foreground colors, caused by reflected
10144 color of greenscreen or bluescreen.
10145 This filter accepts the following options:
10147 type
10149 Set what type of despill to use.
10150 mix Set how spillmap will be generated.
10152 expand
10154 Set how much to get rid of still remaining spill.
10155 red Controls amount of red in spill area.
10157 green
10159 Controls amount of green in spill area. Should be -1 for
10160 greenscreen.
10161 blue
10163 Controls amount of blue in spill area. Should be -1 for
10164 bluescreen.
10165 brightness
10167 Controls brightness of spill area, preserving colors.
10168 alpha
10170 Modify alpha from generated spillmap.
10171 Commands
10173 This filter supports the all above options as commands.
10175 detelecine
10177 Apply an exact inverse of the telecine operation. It requires a
10178 predefined pattern specified using the pattern option which must be the
10179 same as that passed to the telecine filter.
10180 This filter accepts the following options:
10182 first_field
10184 top, t
10185 top field first
10186 bottom, b
10188 bottom field first The default value is "top".
10189 pattern
10191 A string of numbers representing the pulldown pattern you wish to
10192 apply. The default value is 23.
10193 start_frame
10195 A number representing position of the first frame with respect to
10196 the telecine pattern. This is to be used if the stream is cut. The
10197 default value is 0.
10198 dilation
10200 Apply dilation effect to the video.
10201 This filter replaces the pixel by the local(3x3) maximum.
10203 It accepts the following options:
10205 threshold0
10207 threshold1
10208 threshold2
10209 threshold3
10210 Limit the maximum change for each plane, default is 65535. If 0,
10211 plane will remain unchanged.
10212 coordinates
10214 Flag which specifies the pixel to refer to. Default is 255 i.e. all
10215 eight pixels are used.
10216 Flags to local 3x3 coordinates maps like this:
10218 1 2 3
10220 4 5
10221 6 7 8
10222 Commands
10224 This filter supports the all above options as commands.
10226 displace
10228 Displace pixels as indicated by second and third input stream.
10229 It takes three input streams and outputs one stream, the first input is
10231 the source, and second and third input are displacement maps.
10232 The second input specifies how much to displace pixels along the
10234 x-axis, while the third input specifies how much to displace pixels
10235 along the y-axis. If one of displacement map streams terminates, last
10236 frame from that displacement map will be used.
10237 Note that once generated, displacements maps can be reused over and
10239 over again.
10240 A description of the accepted options follows.
10242 edge
10244 Set displace behavior for pixels that are out of range.
10245 Available values are:
10247 blank
10249 Missing pixels are replaced by black pixels.
10250 smear
10252 Adjacent pixels will spread out to replace missing pixels.
10253 wrap
10255 Out of range pixels are wrapped so they point to pixels of
10256 other side.
10257 mirror
10259 Out of range pixels will be replaced with mirrored pixels.
10260 Default is smear.
10262 Examples
10264 • Add ripple effect to rgb input of video size hd720:
10266 ffmpeg -i INPUT -f lavfi -i nullsrc=s=hd720,lutrgb=128:128:128 -f lavfi -i nullsrc=s=hd720,geq='r=128+30*sin(2*PI*X/400+T):g=128+30*sin(2*PI*X/400+T):b=128+30*sin(2*PI*X/400+T)' -lavfi '[0][1][2]displace' OUTPUT
10268 • Add wave effect to rgb input of video size hd720:
10270 ffmpeg -i INPUT -f lavfi -i nullsrc=hd720,geq='r=128+80*(sin(sqrt((X-W/2)*(X-W/2)+(Y-H/2)*(Y-H/2))/220*2*PI+T)):g=128+80*(sin(sqrt((X-W/2)*(X-W/2)+(Y-H/2)*(Y-H/2))/220*2*PI+T)):b=128+80*(sin(sqrt((X-W/2)*(X-W/2)+(Y-H/2)*(Y-H/2))/220*2*PI+T))' -lavfi '[1]split[x][y],[0][x][y]displace' OUTPUT
10272 dnn_classify
10274 Do classification with deep neural networks based on bounding boxes.
10275 The filter accepts the following options:
10277 dnn_backend
10279 Specify which DNN backend to use for model loading and execution.
10280 This option accepts only openvino now, tensorflow backends will be
10281 added.
10282 model
10284 Set path to model file specifying network architecture and its
10285 parameters. Note that different backends use different file
10286 formats.
10287 input
10289 Set the input name of the dnn network.
10290 output
10292 Set the output name of the dnn network.
10293 confidence
10295 Set the confidence threshold (default: 0.5).
10296 labels
10298 Set path to label file specifying the mapping between label id and
10299 name. Each label name is written in one line, tailing spaces and
10300 empty lines are skipped. The first line is the name of label id 0,
10301 and the second line is the name of label id 1, etc. The label id
10302 is considered as name if the label file is not provided.
10303 backend_configs
10305 Set the configs to be passed into backend
10306 For tensorflow backend, you can set its configs with sess_config
10308 options, please use tools/python/tf_sess_config.py to get the
10309 configs for your system.
10310 dnn_detect
10312 Do object detection with deep neural networks.
10313 The filter accepts the following options:
10315 dnn_backend
10317 Specify which DNN backend to use for model loading and execution.
10318 This option accepts only openvino now, tensorflow backends will be
10319 added.
10320 model
10322 Set path to model file specifying network architecture and its
10323 parameters. Note that different backends use different file
10324 formats.
10325 input
10327 Set the input name of the dnn network.
10328 output
10330 Set the output name of the dnn network.
10331 confidence
10333 Set the confidence threshold (default: 0.5).
10334 labels
10336 Set path to label file specifying the mapping between label id and
10337 name. Each label name is written in one line, tailing spaces and
10338 empty lines are skipped. The first line is the name of label id 0
10339 (usually it is 'background'), and the second line is the name of
10340 label id 1, etc. The label id is considered as name if the label
10341 file is not provided.
10342 backend_configs
10344 Set the configs to be passed into backend. To use async execution,
10345 set async (default: set). Roll back to sync execution if the
10346 backend does not support async.
10347 dnn_processing
10349 Do image processing with deep neural networks. It works together with
10350 another filter which converts the pixel format of the Frame to what the
10351 dnn network requires.
10352 The filter accepts the following options:
10354 dnn_backend
10356 Specify which DNN backend to use for model loading and execution.
10357 This option accepts the following values:
10358 native
10360 Native implementation of DNN loading and execution.
10361 tensorflow
10363 TensorFlow backend. To enable this backend you need to install
10364 the TensorFlow for C library (see
10365 <https://www.tensorflow.org/install/lang_c>) and configure
10366 FFmpeg with "--enable-libtensorflow"
10367 openvino
10369 OpenVINO backend. To enable this backend you need to build and
10370 install the OpenVINO for C library (see
10371 <https://github.com/openvinotoolkit/openvino/blob/master/build-instruction.md>)
10372 and configure FFmpeg with "--enable-libopenvino"
10373 (--extra-cflags=-I... --extra-ldflags=-L... might be needed if
10374 the header files and libraries are not installed into system
10375 path)
10376 Default value is native.
10378 model
10380 Set path to model file specifying network architecture and its
10381 parameters. Note that different backends use different file
10382 formats. TensorFlow, OpenVINO and native backend can load files for
10383 only its format.
10384 Native model file (.model) can be generated from TensorFlow model
10386 file (.pb) by using tools/python/convert.py
10387 input
10389 Set the input name of the dnn network.
10390 output
10392 Set the output name of the dnn network.
10393 backend_configs
10395 Set the configs to be passed into backend. To use async execution,
10396 set async (default: set). Roll back to sync execution if the
10397 backend does not support async.
10398 For tensorflow backend, you can set its configs with sess_config
10400 options, please use tools/python/tf_sess_config.py to get the
10401 configs of TensorFlow backend for your system.
10402 Examples
10404 • Remove rain in rgb24 frame with can.pb (see derain filter):
10406 ./ffmpeg -i rain.jpg -vf format=rgb24,dnn_processing=dnn_backend=tensorflow:model=can.pb:input=x:output=y derain.jpg
10408 • Halve the pixel value of the frame with format gray32f:
10410 ffmpeg -i input.jpg -vf format=grayf32,dnn_processing=model=halve_gray_float.model:input=dnn_in:output=dnn_out:dnn_backend=native -y out.native.png
10412 • Handle the Y channel with srcnn.pb (see sr filter) for frame with
10414 yuv420p (planar YUV formats supported):
10415 ./ffmpeg -i 480p.jpg -vf format=yuv420p,scale=w=iw*2:h=ih*2,dnn_processing=dnn_backend=tensorflow:model=srcnn.pb:input=x:output=y -y srcnn.jpg
10417 • Handle the Y channel with espcn.pb (see sr filter), which changes
10419 frame size, for format yuv420p (planar YUV formats supported),
10420 please use tools/python/tf_sess_config.py to get the configs of
10421 TensorFlow backend for your system.
10422 ./ffmpeg -i 480p.jpg -vf format=yuv420p,dnn_processing=dnn_backend=tensorflow:model=espcn.pb:input=x:output=y:backend_configs=sess_config=0x10022805320e09cdccccccccccec3f20012a01303801 -y tmp.espcn.jpg
10424 drawbox
10426 Draw a colored box on the input image.
10427 It accepts the following parameters:
10429 x
10431 y The expressions which specify the top left corner coordinates of
10432 the box. It defaults to 0.
10433 width, w
10435 height, h
10436 The expressions which specify the width and height of the box; if 0
10437 they are interpreted as the input width and height. It defaults to
10438 0.
10439 color, c
10441 Specify the color of the box to write. For the general syntax of
10442 this option, check the "Color" section in the ffmpeg-utils manual.
10443 If the special value "invert" is used, the box edge color is the
10444 same as the video with inverted luma.
10445 thickness, t
10447 The expression which sets the thickness of the box edge. A value
10448 of "fill" will create a filled box. Default value is 3.
10449 See below for the list of accepted constants.
10451 replace
10453 Applicable if the input has alpha. With value 1, the pixels of the
10454 painted box will overwrite the video's color and alpha pixels.
10455 Default is 0, which composites the box onto the input, leaving the
10456 video's alpha intact.
10457 The parameters for x, y, w and h and t are expressions containing the
10459 following constants:
10460 dar The input display aspect ratio, it is the same as (w / h) * sar.
10462 hsub
10464 vsub
10465 horizontal and vertical chroma subsample values. For example for
10466 the pixel format "yuv422p" hsub is 2 and vsub is 1.
10467 in_h, ih
10469 in_w, iw
10470 The input width and height.
10471 sar The input sample aspect ratio.
10473 x
10475 y The x and y offset coordinates where the box is drawn.
10476 w
10478 h The width and height of the drawn box.
10479 box_source
10481 Box source can be set as side_data_detection_bboxes if you want to
10482 use box data in detection bboxes of side data.
10483 If box_source is set, the x, y, width and height will be ignored
10485 and still use box data in detection bboxes of side data. So please
10486 do not use this parameter if you were not sure about the box
10487 source.
10488 t The thickness of the drawn box.
10490 These constants allow the x, y, w, h and t expressions to refer to
10492 each other, so you may for example specify "y=x/dar" or "h=w/dar".
10493 Examples
10495 • Draw a black box around the edge of the input image:
10497 drawbox
10499 • Draw a box with color red and an opacity of 50%:
10501 drawbox=10:20:200:60:red@0.5
10503 The previous example can be specified as:
10505 drawbox=x=10:y=20:w=200:h=60:color=red@0.5
10507 • Fill the box with pink color:
10509 drawbox=x=10:y=10:w=100:h=100:color=pink@0.5:t=fill
10511 • Draw a 2-pixel red 2.40:1 mask:
10513 drawbox=x=-t:y=0.5*(ih-iw/2.4)-t:w=iw+t*2:h=iw/2.4+t*2:t=2:c=red
10515 Commands
10517 This filter supports same commands as options. The command accepts the
10519 same syntax of the corresponding option.
10520 If the specified expression is not valid, it is kept at its current
10522 value.
10523 drawgraph
10525 Draw a graph using input video metadata.
10526 It accepts the following parameters:
10528 m1 Set 1st frame metadata key from which metadata values will be used
10530 to draw a graph.
10531 fg1 Set 1st foreground color expression.
10533 m2 Set 2nd frame metadata key from which metadata values will be used
10535 to draw a graph.
10536 fg2 Set 2nd foreground color expression.
10538 m3 Set 3rd frame metadata key from which metadata values will be used
10540 to draw a graph.
10541 fg3 Set 3rd foreground color expression.
10543 m4 Set 4th frame metadata key from which metadata values will be used
10545 to draw a graph.
10546 fg4 Set 4th foreground color expression.
10548 min Set minimal value of metadata value.
10550 max Set maximal value of metadata value.
10552 bg Set graph background color. Default is white.
10554 mode
10556 Set graph mode.
10557 Available values for mode is:
10559 bar
10561 dot
10562 line
10563 Default is "line".
10565 slide
10567 Set slide mode.
10568 Available values for slide is:
10570 frame
10572 Draw new frame when right border is reached.
10573 replace
10575 Replace old columns with new ones.
10576 scroll
10578 Scroll from right to left.
10579 rscroll
10581 Scroll from left to right.
10582 picture
10584 Draw single picture.
10585 Default is "frame".
10587 size
10589 Set size of graph video. For the syntax of this option, check the
10590 "Video size" section in the ffmpeg-utils manual. The default value
10591 is "900x256".
10592 rate, r
10594 Set the output frame rate. Default value is 25.
10595 The foreground color expressions can use the following variables:
10597 MIN Minimal value of metadata value.
10599 MAX Maximal value of metadata value.
10601 VAL Current metadata key value.
10603 The color is defined as 0xAABBGGRR.
10605 Example using metadata from signalstats filter:
10607 signalstats,drawgraph=lavfi.signalstats.YAVG:min=0:max=255
10609 Example using metadata from ebur128 filter:
10611 ebur128=metadata=1,adrawgraph=lavfi.r128.M:min=-120:max=5
10613 drawgrid
10615 Draw a grid on the input image.
10616 It accepts the following parameters:
10618 x
10620 y The expressions which specify the coordinates of some point of grid
10621 intersection (meant to configure offset). Both default to 0.
10622 width, w
10624 height, h
10625 The expressions which specify the width and height of the grid
10626 cell, if 0 they are interpreted as the input width and height,
10627 respectively, minus "thickness", so image gets framed. Default to
10628 0.
10629 color, c
10631 Specify the color of the grid. For the general syntax of this
10632 option, check the "Color" section in the ffmpeg-utils manual. If
10633 the special value "invert" is used, the grid color is the same as
10634 the video with inverted luma.
10635 thickness, t
10637 The expression which sets the thickness of the grid line. Default
10638 value is 1.
10639 See below for the list of accepted constants.
10641 replace
10643 Applicable if the input has alpha. With 1 the pixels of the painted
10644 grid will overwrite the video's color and alpha pixels. Default is
10645 0, which composites the grid onto the input, leaving the video's
10646 alpha intact.
10647 The parameters for x, y, w and h and t are expressions containing the
10649 following constants:
10650 dar The input display aspect ratio, it is the same as (w / h) * sar.
10652 hsub
10654 vsub
10655 horizontal and vertical chroma subsample values. For example for
10656 the pixel format "yuv422p" hsub is 2 and vsub is 1.
10657 in_h, ih
10659 in_w, iw
10660 The input grid cell width and height.
10661 sar The input sample aspect ratio.
10663 x
10665 y The x and y coordinates of some point of grid intersection (meant
10666 to configure offset).
10667 w
10669 h The width and height of the drawn cell.
10670 t The thickness of the drawn cell.
10672 These constants allow the x, y, w, h and t expressions to refer to
10674 each other, so you may for example specify "y=x/dar" or "h=w/dar".
10675 Examples
10677 • Draw a grid with cell 100x100 pixels, thickness 2 pixels, with
10679 color red and an opacity of 50%:
10680 drawgrid=width=100:height=100:thickness=2:color=red@0.5
10682 • Draw a white 3x3 grid with an opacity of 50%:
10684 drawgrid=w=iw/3:h=ih/3:t=2:c=white@0.5
10686 Commands
10688 This filter supports same commands as options. The command accepts the
10690 same syntax of the corresponding option.
10691 If the specified expression is not valid, it is kept at its current
10693 value.
10694 drawtext
10696 Draw a text string or text from a specified file on top of a video,
10697 using the libfreetype library.
10698 To enable compilation of this filter, you need to configure FFmpeg with
10700 "--enable-libfreetype". To enable default font fallback and the font
10701 option you need to configure FFmpeg with "--enable-libfontconfig". To
10702 enable the text_shaping option, you need to configure FFmpeg with
10703 "--enable-libfribidi".
10704 Syntax
10706 It accepts the following parameters:
10708 box Used to draw a box around text using the background color. The
10710 value must be either 1 (enable) or 0 (disable). The default value
10711 of box is 0.
10712 boxborderw
10714 Set the width of the border to be drawn around the box using
10715 boxcolor. The default value of boxborderw is 0.
10716 boxcolor
10718 The color to be used for drawing box around text. For the syntax of
10719 this option, check the "Color" section in the ffmpeg-utils manual.
10720 The default value of boxcolor is "white".
10722 line_spacing
10724 Set the line spacing in pixels of the border to be drawn around the
10725 box using box. The default value of line_spacing is 0.
10726 borderw
10728 Set the width of the border to be drawn around the text using
10729 bordercolor. The default value of borderw is 0.
10730 bordercolor
10732 Set the color to be used for drawing border around text. For the
10733 syntax of this option, check the "Color" section in the ffmpeg-
10734 utils manual.
10735 The default value of bordercolor is "black".
10737 expansion
10739 Select how the text is expanded. Can be either "none", "strftime"
10740 (deprecated) or "normal" (default). See the drawtext_expansion,
10741 Text expansion section below for details.
10742 basetime
10744 Set a start time for the count. Value is in microseconds. Only
10745 applied in the deprecated strftime expansion mode. To emulate in
10746 normal expansion mode use the "pts" function, supplying the start
10747 time (in seconds) as the second argument.
10748 fix_bounds
10750 If true, check and fix text coords to avoid clipping.
10751 fontcolor
10753 The color to be used for drawing fonts. For the syntax of this
10754 option, check the "Color" section in the ffmpeg-utils manual.
10755 The default value of fontcolor is "black".
10757 fontcolor_expr
10759 String which is expanded the same way as text to obtain dynamic
10760 fontcolor value. By default this option has empty value and is not
10761 processed. When this option is set, it overrides fontcolor option.
10762 font
10764 The font family to be used for drawing text. By default Sans.
10765 fontfile
10767 The font file to be used for drawing text. The path must be
10768 included. This parameter is mandatory if the fontconfig support is
10769 disabled.
10770 alpha
10772 Draw the text applying alpha blending. The value can be a number
10773 between 0.0 and 1.0. The expression accepts the same variables x,
10774 y as well. The default value is 1. Please see fontcolor_expr.
10775 fontsize
10777 The font size to be used for drawing text. The default value of
10778 fontsize is 16.
10779 text_shaping
10781 If set to 1, attempt to shape the text (for example, reverse the
10782 order of right-to-left text and join Arabic characters) before
10783 drawing it. Otherwise, just draw the text exactly as given. By
10784 default 1 (if supported).
10785 ft_load_flags
10787 The flags to be used for loading the fonts.
10788 The flags map the corresponding flags supported by libfreetype, and
10790 are a combination of the following values:
10791 default
10793 no_scale
10794 no_hinting
10795 render
10796 no_bitmap
10797 vertical_layout
10798 force_autohint
10799 crop_bitmap
10800 pedantic
10801 ignore_global_advance_width
10802 no_recurse
10803 ignore_transform
10804 monochrome
10805 linear_design
10806 no_autohint
10807 Default value is "default".
10809 For more information consult the documentation for the FT_LOAD_*
10811 libfreetype flags.
10812 shadowcolor
10814 The color to be used for drawing a shadow behind the drawn text.
10815 For the syntax of this option, check the "Color" section in the
10816 ffmpeg-utils manual.
10817 The default value of shadowcolor is "black".
10819 shadowx
10821 shadowy
10822 The x and y offsets for the text shadow position with respect to
10823 the position of the text. They can be either positive or negative
10824 values. The default value for both is "0".
10825 start_number
10827 The starting frame number for the n/frame_num variable. The default
10828 value is "0".
10829 tabsize
10831 The size in number of spaces to use for rendering the tab. Default
10832 value is 4.
10833 timecode
10835 Set the initial timecode representation in "hh:mm:ss[:;.]ff"
10836 format. It can be used with or without text parameter.
10837 timecode_rate option must be specified.
10838 timecode_rate, rate, r
10840 Set the timecode frame rate (timecode only). Value will be rounded
10841 to nearest integer. Minimum value is "1". Drop-frame timecode is
10842 supported for frame rates 30 & 60.
10843 tc24hmax
10845 If set to 1, the output of the timecode option will wrap around at
10846 24 hours. Default is 0 (disabled).
10847 text
10849 The text string to be drawn. The text must be a sequence of UTF-8
10850 encoded characters. This parameter is mandatory if no file is
10851 specified with the parameter textfile.
10852 textfile
10854 A text file containing text to be drawn. The text must be a
10855 sequence of UTF-8 encoded characters.
10856 This parameter is mandatory if no text string is specified with the
10858 parameter text.
10859 If both text and textfile are specified, an error is thrown.
10861 text_source
10863 Text source should be set as side_data_detection_bboxes if you want
10864 to use text data in detection bboxes of side data.
10865 If text source is set, text and textfile will be ignored and still
10867 use text data in detection bboxes of side data. So please do not
10868 use this parameter if you are not sure about the text source.
10869 reload
10871 The textfile will be reloaded at specified frame interval. Be sure
10872 to update textfile atomically, or it may be read partially, or even
10873 fail. Range is 0 to INT_MAX. Default is 0.
10874 x
10876 y The expressions which specify the offsets where text will be drawn
10877 within the video frame. They are relative to the top/left border of
10878 the output image.
10879 The default value of x and y is "0".
10881 See below for the list of accepted constants and functions.
10883 The parameters for x and y are expressions containing the following
10885 constants and functions:
10886 dar input display aspect ratio, it is the same as (w / h) * sar
10888 hsub
10890 vsub
10891 horizontal and vertical chroma subsample values. For example for
10892 the pixel format "yuv422p" hsub is 2 and vsub is 1.
10893 line_h, lh
10895 the height of each text line
10896 main_h, h, H
10898 the input height
10899 main_w, w, W
10901 the input width
10902 max_glyph_a, ascent
10904 the maximum distance from the baseline to the highest/upper grid
10905 coordinate used to place a glyph outline point, for all the
10906 rendered glyphs. It is a positive value, due to the grid's
10907 orientation with the Y axis upwards.
10908 max_glyph_d, descent
10910 the maximum distance from the baseline to the lowest grid
10911 coordinate used to place a glyph outline point, for all the
10912 rendered glyphs. This is a negative value, due to the grid's
10913 orientation, with the Y axis upwards.
10914 max_glyph_h
10916 maximum glyph height, that is the maximum height for all the glyphs
10917 contained in the rendered text, it is equivalent to ascent -
10918 descent.
10919 max_glyph_w
10921 maximum glyph width, that is the maximum width for all the glyphs
10922 contained in the rendered text
10923 n the number of input frame, starting from 0
10925 rand(min, max)
10927 return a random number included between min and max
10928 sar The input sample aspect ratio.
10930 t timestamp expressed in seconds, NAN if the input timestamp is
10932 unknown
10933 text_h, th
10935 the height of the rendered text
10936 text_w, tw
10938 the width of the rendered text
10939 x
10941 y the x and y offset coordinates where the text is drawn.
10942 These parameters allow the x and y expressions to refer to each
10944 other, so you can for example specify "y=x/dar".
10945 pict_type
10947 A one character description of the current frame's picture type.
10948 pkt_pos
10950 The current packet's position in the input file or stream (in
10951 bytes, from the start of the input). A value of -1 indicates this
10952 info is not available.
10953 duration
10955 The current packet's duration, in seconds.
10956 pkt_size
10958 The current packet's size (in bytes).
10959 Text expansion
10961 If expansion is set to "strftime", the filter recognizes strftime()
10963 sequences in the provided text and expands them accordingly. Check the
10964 documentation of strftime(). This feature is deprecated.
10965 If expansion is set to "none", the text is printed verbatim.
10967 If expansion is set to "normal" (which is the default), the following
10969 expansion mechanism is used.
10970 The backslash character \, followed by any character, always expands to
10972 the second character.
10973 Sequences of the form "%{...}" are expanded. The text between the
10975 braces is a function name, possibly followed by arguments separated by
10976 ':'. If the arguments contain special characters or delimiters (':' or
10977 '}'), they should be escaped.
10978 Note that they probably must also be escaped as the value for the text
10980 option in the filter argument string and as the filter argument in the
10981 filtergraph description, and possibly also for the shell, that makes up
10982 to four levels of escaping; using a text file avoids these problems.
10983 The following functions are available:
10985 expr, e
10987 The expression evaluation result.
10988 It must take one argument specifying the expression to be
10990 evaluated, which accepts the same constants and functions as the x
10991 and y values. Note that not all constants should be used, for
10992 example the text size is not known when evaluating the expression,
10993 so the constants text_w and text_h will have an undefined value.
10994 expr_int_format, eif
10996 Evaluate the expression's value and output as formatted integer.
10997 The first argument is the expression to be evaluated, just as for
10999 the expr function. The second argument specifies the output
11000 format. Allowed values are x, X, d and u. They are treated exactly
11001 as in the "printf" function. The third parameter is optional and
11002 sets the number of positions taken by the output. It can be used
11003 to add padding with zeros from the left.
11004 gmtime
11006 The time at which the filter is running, expressed in UTC. It can
11007 accept an argument: a strftime() format string. The format string
11008 is extended to support the variable %[1-6]N which prints fractions
11009 of the second with optionally specified number of digits.
11010 localtime
11012 The time at which the filter is running, expressed in the local
11013 time zone. It can accept an argument: a strftime() format string.
11014 The format string is extended to support the variable %[1-6]N which
11015 prints fractions of the second with optionally specified number of
11016 digits.
11017 metadata
11019 Frame metadata. Takes one or two arguments.
11020 The first argument is mandatory and specifies the metadata key.
11022 The second argument is optional and specifies a default value, used
11024 when the metadata key is not found or empty.
11025 Available metadata can be identified by inspecting entries starting
11027 with TAG included within each frame section printed by running
11028 "ffprobe -show_frames".
11029 String metadata generated in filters leading to the drawtext filter
11031 are also available.
11032 n, frame_num
11034 The frame number, starting from 0.
11035 pict_type
11037 A one character description of the current picture type.
11038 pts The timestamp of the current frame. It can take up to three
11040 arguments.
11041 The first argument is the format of the timestamp; it defaults to
11043 "flt" for seconds as a decimal number with microsecond accuracy;
11044 "hms" stands for a formatted [-]HH:MM:SS.mmm timestamp with
11045 millisecond accuracy. "gmtime" stands for the timestamp of the
11046 frame formatted as UTC time; "localtime" stands for the timestamp
11047 of the frame formatted as local time zone time.
11048 The second argument is an offset added to the timestamp.
11050 If the format is set to "hms", a third argument "24HH" may be
11052 supplied to present the hour part of the formatted timestamp in 24h
11053 format (00-23).
11054 If the format is set to "localtime" or "gmtime", a third argument
11056 may be supplied: a strftime() format string. By default, YYYY-MM-
11057 DD HH:MM:SS format will be used.
11058 Commands
11060 This filter supports altering parameters via commands:
11062 reinit
11064 Alter existing filter parameters.
11065 Syntax for the argument is the same as for filter invocation, e.g.
11067 fontsize=56:fontcolor=green:text='Hello World'
11069 Full filter invocation with sendcmd would look like this:
11071 sendcmd=c='56.0 drawtext reinit fontsize=56\:fontcolor=green\:text=Hello\\ World'
11073 If the entire argument can't be parsed or applied as valid values then
11075 the filter will continue with its existing parameters.
11076 Examples
11078 • Draw "Test Text" with font FreeSerif, using the default values for
11080 the optional parameters.
11081 drawtext="fontfile=/usr/share/fonts/truetype/freefont/FreeSerif.ttf: text='Test Text'"
11083 • Draw 'Test Text' with font FreeSerif of size 24 at position x=100
11085 and y=50 (counting from the top-left corner of the screen), text is
11086 yellow with a red box around it. Both the text and the box have an
11087 opacity of 20%.
11088 drawtext="fontfile=/usr/share/fonts/truetype/freefont/FreeSerif.ttf: text='Test Text':\
11090 x=100: y=50: fontsize=24: fontcolor=yellow@0.2: box=1: boxcolor=red@0.2"
11091 Note that the double quotes are not necessary if spaces are not
11093 used within the parameter list.
11094 • Show the text at the center of the video frame:
11096 drawtext="fontsize=30:fontfile=FreeSerif.ttf:text='hello world':x=(w-text_w)/2:y=(h-text_h)/2"
11098 • Show the text at a random position, switching to a new position
11100 every 30 seconds:
11101 drawtext="fontsize=30:fontfile=FreeSerif.ttf:text='hello world':x=if(eq(mod(t\,30)\,0)\,rand(0\,(w-text_w))\,x):y=if(eq(mod(t\,30)\,0)\,rand(0\,(h-text_h))\,y)"
11103 • Show a text line sliding from right to left in the last row of the
11105 video frame. The file LONG_LINE is assumed to contain a single line
11106 with no newlines.
11107 drawtext="fontsize=15:fontfile=FreeSerif.ttf:text=LONG_LINE:y=h-line_h:x=-50*t"
11109 • Show the content of file CREDITS off the bottom of the frame and
11111 scroll up.
11112 drawtext="fontsize=20:fontfile=FreeSerif.ttf:textfile=CREDITS:y=h-20*t"
11114 • Draw a single green letter "g", at the center of the input video.
11116 The glyph baseline is placed at half screen height.
11117 drawtext="fontsize=60:fontfile=FreeSerif.ttf:fontcolor=green:text=g:x=(w-max_glyph_w)/2:y=h/2-ascent"
11119 • Show text for 1 second every 3 seconds:
11121 drawtext="fontfile=FreeSerif.ttf:fontcolor=white:x=100:y=x/dar:enable=lt(mod(t\,3)\,1):text='blink'"
11123 • Use fontconfig to set the font. Note that the colons need to be
11125 escaped.
11126 drawtext='fontfile=Linux Libertine O-40\:style=Semibold:text=FFmpeg'
11128 • Draw "Test Text" with font size dependent on height of the video.
11130 drawtext="text='Test Text': fontsize=h/30: x=(w-text_w)/2: y=(h-text_h*2)"
11132 • Print the date of a real-time encoding (see strftime(3)):
11134 drawtext='fontfile=FreeSans.ttf:text=%{localtime\:%a %b %d %Y}'
11136 • Show text fading in and out (appearing/disappearing):
11138 #!/bin/sh
11140 DS=1.0 # display start
11141 DE=10.0 # display end
11142 FID=1.5 # fade in duration
11143 FOD=5 # fade out duration
11144 ffplay -f lavfi "color,drawtext=text=TEST:fontsize=50:fontfile=FreeSerif.ttf:fontcolor_expr=ff0000%{eif\\\\: clip(255*(1*between(t\\, $DS + $FID\\, $DE - $FOD) + ((t - $DS)/$FID)*between(t\\, $DS\\, $DS + $FID) + (-(t - $DE)/$FOD)*between(t\\, $DE - $FOD\\, $DE) )\\, 0\\, 255) \\\\: x\\\\: 2 }"
11145 • Horizontally align multiple separate texts. Note that max_glyph_a
11147 and the fontsize value are included in the y offset.
11148 drawtext=fontfile=FreeSans.ttf:text=DOG:fontsize=24:x=10:y=20+24-max_glyph_a,
11150 drawtext=fontfile=FreeSans.ttf:text=cow:fontsize=24:x=80:y=20+24-max_glyph_a
11151 • Plot special lavf.image2dec.source_basename metadata onto each
11153 frame if such metadata exists. Otherwise, plot the string "NA".
11154 Note that image2 demuxer must have option -export_path_metadata 1
11155 for the special metadata fields to be available for filters.
11156 drawtext="fontsize=20:fontcolor=white:fontfile=FreeSans.ttf:text='%{metadata\:lavf.image2dec.source_basename\:NA}':x=10:y=10"
11158 For more information about libfreetype, check:
11160 <http://www.freetype.org/>.
11161 For more information about fontconfig, check:
11163 <http://freedesktop.org/software/fontconfig/fontconfig-user.html>.
11164 For more information about libfribidi, check: <http://fribidi.org/>.
11166 edgedetect
11168 Detect and draw edges. The filter uses the Canny Edge Detection
11169 algorithm.
11170 The filter accepts the following options:
11172 low
11174 high
11175 Set low and high threshold values used by the Canny thresholding
11176 algorithm.
11177 The high threshold selects the "strong" edge pixels, which are then
11179 connected through 8-connectivity with the "weak" edge pixels
11180 selected by the low threshold.
11181 low and high threshold values must be chosen in the range [0,1],
11183 and low should be lesser or equal to high.
11184 Default value for low is "20/255", and default value for high is
11186 "50/255".
11187 mode
11189 Define the drawing mode.
11190 wires
11192 Draw white/gray wires on black background.
11193 colormix
11195 Mix the colors to create a paint/cartoon effect.
11196 canny
11198 Apply Canny edge detector on all selected planes.
11199 Default value is wires.
11201 planes
11203 Select planes for filtering. By default all available planes are
11204 filtered.
11205 Examples
11207 • Standard edge detection with custom values for the hysteresis
11209 thresholding:
11210 edgedetect=low=0.1:high=0.4
11212 • Painting effect without thresholding:
11214 edgedetect=mode=colormix:high=0
11216 elbg
11218 Apply a posterize effect using the ELBG (Enhanced LBG) algorithm.
11219 For each input image, the filter will compute the optimal mapping from
11221 the input to the output given the codebook length, that is the number
11222 of distinct output colors.
11223 This filter accepts the following options.
11225 codebook_length, l
11227 Set codebook length. The value must be a positive integer, and
11228 represents the number of distinct output colors. Default value is
11229 256.
11230 nb_steps, n
11232 Set the maximum number of iterations to apply for computing the
11233 optimal mapping. The higher the value the better the result and the
11234 higher the computation time. Default value is 1.
11235 seed, s
11237 Set a random seed, must be an integer included between 0 and
11238 UINT32_MAX. If not specified, or if explicitly set to -1, the
11239 filter will try to use a good random seed on a best effort basis.
11240 pal8
11242 Set pal8 output pixel format. This option does not work with
11243 codebook length greater than 256. Default is disabled.
11244 use_alpha
11246 Include alpha values in the quantization calculation. Allows
11247 creating palettized output images (e.g. PNG8) with multiple alpha
11248 smooth blending.
11249 entropy
11251 Measure graylevel entropy in histogram of color channels of video
11252 frames.
11253 It accepts the following parameters:
11255 mode
11257 Can be either normal or diff. Default is normal.
11258 diff mode measures entropy of histogram delta values, absolute
11260 differences between neighbour histogram values.
11261 epx
11263 Apply the EPX magnification filter which is designed for pixel art.
11264 It accepts the following option:
11266 n Set the scaling dimension: 2 for "2xEPX", 3 for "3xEPX". Default
11268 is 3.
11269 eq
11271 Set brightness, contrast, saturation and approximate gamma adjustment.
11272 The filter accepts the following options:
11274 contrast
11276 Set the contrast expression. The value must be a float value in
11277 range -1000.0 to 1000.0. The default value is "1".
11278 brightness
11280 Set the brightness expression. The value must be a float value in
11281 range -1.0 to 1.0. The default value is "0".
11282 saturation
11284 Set the saturation expression. The value must be a float in range
11285 0.0 to 3.0. The default value is "1".
11286 gamma
11288 Set the gamma expression. The value must be a float in range 0.1 to
11289 10.0. The default value is "1".
11290 gamma_r
11292 Set the gamma expression for red. The value must be a float in
11293 range 0.1 to 10.0. The default value is "1".
11294 gamma_g
11296 Set the gamma expression for green. The value must be a float in
11297 range 0.1 to 10.0. The default value is "1".
11298 gamma_b
11300 Set the gamma expression for blue. The value must be a float in
11301 range 0.1 to 10.0. The default value is "1".
11302 gamma_weight
11304 Set the gamma weight expression. It can be used to reduce the
11305 effect of a high gamma value on bright image areas, e.g. keep them
11306 from getting overamplified and just plain white. The value must be
11307 a float in range 0.0 to 1.0. A value of 0.0 turns the gamma
11308 correction all the way down while 1.0 leaves it at its full
11309 strength. Default is "1".
11310 eval
11312 Set when the expressions for brightness, contrast, saturation and
11313 gamma expressions are evaluated.
11314 It accepts the following values:
11316 init
11318 only evaluate expressions once during the filter initialization
11319 or when a command is processed
11320 frame
11322 evaluate expressions for each incoming frame
11323 Default value is init.
11325 The expressions accept the following parameters:
11327 n frame count of the input frame starting from 0
11329 pos byte position of the corresponding packet in the input file, NAN if
11331 unspecified
11332 r frame rate of the input video, NAN if the input frame rate is
11334 unknown
11335 t timestamp expressed in seconds, NAN if the input timestamp is
11337 unknown
11338 Commands
11340 The filter supports the following commands:
11342 contrast
11344 Set the contrast expression.
11345 brightness
11347 Set the brightness expression.
11348 saturation
11350 Set the saturation expression.
11351 gamma
11353 Set the gamma expression.
11354 gamma_r
11356 Set the gamma_r expression.
11357 gamma_g
11359 Set gamma_g expression.
11360 gamma_b
11362 Set gamma_b expression.
11363 gamma_weight
11365 Set gamma_weight expression.
11366 The command accepts the same syntax of the corresponding option.
11368 If the specified expression is not valid, it is kept at its current
11370 value.
11371 erosion
11373 Apply erosion effect to the video.
11374 This filter replaces the pixel by the local(3x3) minimum.
11376 It accepts the following options:
11378 threshold0
11380 threshold1
11381 threshold2
11382 threshold3
11383 Limit the maximum change for each plane, default is 65535. If 0,
11384 plane will remain unchanged.
11385 coordinates
11387 Flag which specifies the pixel to refer to. Default is 255 i.e. all
11388 eight pixels are used.
11389 Flags to local 3x3 coordinates maps like this:
11391 1 2 3
11393 4 5
11394 6 7 8
11395 Commands
11397 This filter supports the all above options as commands.
11399 estdif
11401 Deinterlace the input video ("estdif" stands for "Edge Slope Tracing
11402 Deinterlacing Filter").
11403 Spatial only filter that uses edge slope tracing algorithm to
11405 interpolate missing lines. It accepts the following parameters:
11406 mode
11408 The interlacing mode to adopt. It accepts one of the following
11409 values:
11410 frame
11412 Output one frame for each frame.
11413 field
11415 Output one frame for each field.
11416 The default value is "field".
11418 parity
11420 The picture field parity assumed for the input interlaced video. It
11421 accepts one of the following values:
11422 tff Assume the top field is first.
11424 bff Assume the bottom field is first.
11426 auto
11428 Enable automatic detection of field parity.
11429 The default value is "auto". If the interlacing is unknown or the
11431 decoder does not export this information, top field first will be
11432 assumed.
11433 deint
11435 Specify which frames to deinterlace. Accepts one of the following
11436 values:
11437 all Deinterlace all frames.
11439 interlaced
11441 Only deinterlace frames marked as interlaced.
11442 The default value is "all".
11444 rslope
11446 Specify the search radius for edge slope tracing. Default value is
11447 1. Allowed range is from 1 to 15.
11448 redge
11450 Specify the search radius for best edge matching. Default value is
11451 2. Allowed range is from 0 to 15.
11452 ecost
11454 Specify the edge cost for edge matching. Default value is 1.0.
11455 Allowed range is from 0 to 9.
11456 mcost
11458 Specify the middle cost for edge matching. Default value is 0.5.
11459 Allowed range is from 0 to 1.
11460 dcost
11462 Specify the distance cost for edge matching. Default value is 0.5.
11463 Allowed range is from 0 to 1.
11464 interp
11466 Specify the interpolation used. Default is 4-point interpolation.
11467 It accepts one of the following values:
11468 2p Two-point interpolation.
11470 4p Four-point interpolation.
11472 6p Six-point interpolation.
11474 Commands
11476 This filter supports same commands as options.
11478 exposure
11480 Adjust exposure of the video stream.
11481 The filter accepts the following options:
11483 exposure
11485 Set the exposure correction in EV. Allowed range is from -3.0 to
11486 3.0 EV Default value is 0 EV.
11487 black
11489 Set the black level correction. Allowed range is from -1.0 to 1.0.
11490 Default value is 0.
11491 Commands
11493 This filter supports same commands as options.
11495 extractplanes
11497 Extract color channel components from input video stream into separate
11498 grayscale video streams.
11499 The filter accepts the following option:
11501 planes
11503 Set plane(s) to extract.
11504 Available values for planes are:
11506 y
11508 u
11509 v
11510 a
11511 r
11512 g
11513 b
11514 Choosing planes not available in the input will result in an error.
11516 That means you cannot select "r", "g", "b" planes with "y", "u",
11517 "v" planes at same time.
11518 Examples
11520 • Extract luma, u and v color channel component from input video
11522 frame into 3 grayscale outputs:
11523 ffmpeg -i video.avi -filter_complex 'extractplanes=y+u+v[y][u][v]' -map '[y]' y.avi -map '[u]' u.avi -map '[v]' v.avi
11525 fade
11527 Apply a fade-in/out effect to the input video.
11528 It accepts the following parameters:
11530 type, t
11532 The effect type can be either "in" for a fade-in, or "out" for a
11533 fade-out effect. Default is "in".
11534 start_frame, s
11536 Specify the number of the frame to start applying the fade effect
11537 at. Default is 0.
11538 nb_frames, n
11540 The number of frames that the fade effect lasts. At the end of the
11541 fade-in effect, the output video will have the same intensity as
11542 the input video. At the end of the fade-out transition, the output
11543 video will be filled with the selected color. Default is 25.
11544 alpha
11546 If set to 1, fade only alpha channel, if one exists on the input.
11547 Default value is 0.
11548 start_time, st
11550 Specify the timestamp (in seconds) of the frame to start to apply
11551 the fade effect. If both start_frame and start_time are specified,
11552 the fade will start at whichever comes last. Default is 0.
11553 duration, d
11555 The number of seconds for which the fade effect has to last. At the
11556 end of the fade-in effect the output video will have the same
11557 intensity as the input video, at the end of the fade-out transition
11558 the output video will be filled with the selected color. If both
11559 duration and nb_frames are specified, duration is used. Default is
11560 0 (nb_frames is used by default).
11561 color, c
11563 Specify the color of the fade. Default is "black".
11564 Examples
11566 • Fade in the first 30 frames of video:
11568 fade=in:0:30
11570 The command above is equivalent to:
11572 fade=t=in:s=0:n=30
11574 • Fade out the last 45 frames of a 200-frame video:
11576 fade=out:155:45
11578 fade=type=out:start_frame=155:nb_frames=45
11579 • Fade in the first 25 frames and fade out the last 25 frames of a
11581 1000-frame video:
11582 fade=in:0:25, fade=out:975:25
11584 • Make the first 5 frames yellow, then fade in from frame 5-24:
11586 fade=in:5:20:color=yellow
11588 • Fade in alpha over first 25 frames of video:
11590 fade=in:0:25:alpha=1
11592 • Make the first 5.5 seconds black, then fade in for 0.5 seconds:
11594 fade=t=in:st=5.5:d=0.5
11596 feedback
11598 Apply feedback video filter.
11599 This filter pass cropped input frames to 2nd output. From there it can
11601 be filtered with other video filters. After filter receives frame from
11602 2nd input, that frame is combined on top of original frame from 1st
11603 input and passed to 1st output.
11604 The typical usage is filter only part of frame.
11606 The filter accepts the following options:
11608 x
11610 y Set the top left crop position.
11611 w
11613 h Set the crop size.
11614 Examples
11616 • Blur only top left rectangular part of video frame size 100x100
11618 with gblur filter.
11619 [in][blurin]feedback=x=0:y=0:w=100:h=100[out][blurout];[blurout]gblur=8[blurin]
11621 • Draw black box on top left part of video frame of size 100x100 with
11623 drawbox filter.
11624 [in][blurin]feedback=x=0:y=0:w=100:h=100[out][blurout];[blurout]drawbox=x=0:y=0:w=100:h=100:t=100[blurin]
11626 fftdnoiz
11628 Denoise frames using 3D FFT (frequency domain filtering).
11629 The filter accepts the following options:
11631 sigma
11633 Set the noise sigma constant. This sets denoising strength.
11634 Default value is 1. Allowed range is from 0 to 30. Using very high
11635 sigma with low overlap may give blocking artifacts.
11636 amount
11638 Set amount of denoising. By default all detected noise is reduced.
11639 Default value is 1. Allowed range is from 0 to 1.
11640 block
11642 Set size of block in pixels, Default is 32, can be 8 to 256.
11643 overlap
11645 Set block overlap. Default is 0.5. Allowed range is from 0.2 to
11646 0.8.
11647 method
11649 Set denoising method. Default is "wiener", can also be "hard".
11650 prev
11652 Set number of previous frames to use for denoising. By default is
11653 set to 0.
11654 next
11656 Set number of next frames to to use for denoising. By default is
11657 set to 0.
11658 planes
11660 Set planes which will be filtered, by default are all available
11661 filtered except alpha.
11662 fftfilt
11664 Apply arbitrary expressions to samples in frequency domain
11665 dc_Y
11667 Adjust the dc value (gain) of the luma plane of the image. The
11668 filter accepts an integer value in range 0 to 1000. The default
11669 value is set to 0.
11670 dc_U
11672 Adjust the dc value (gain) of the 1st chroma plane of the image.
11673 The filter accepts an integer value in range 0 to 1000. The default
11674 value is set to 0.
11675 dc_V
11677 Adjust the dc value (gain) of the 2nd chroma plane of the image.
11678 The filter accepts an integer value in range 0 to 1000. The default
11679 value is set to 0.
11680 weight_Y
11682 Set the frequency domain weight expression for the luma plane.
11683 weight_U
11685 Set the frequency domain weight expression for the 1st chroma
11686 plane.
11687 weight_V
11689 Set the frequency domain weight expression for the 2nd chroma
11690 plane.
11691 eval
11693 Set when the expressions are evaluated.
11694 It accepts the following values:
11696 init
11698 Only evaluate expressions once during the filter
11699 initialization.
11700 frame
11702 Evaluate expressions for each incoming frame.
11703 Default value is init.
11705 The filter accepts the following variables:
11707 X
11709 Y The coordinates of the current sample.
11710 W
11712 H The width and height of the image.
11713 N The number of input frame, starting from 0.
11715 WS
11717 HS The size of FFT array for horizontal and vertical processing.
11718 Examples
11720 • High-pass:
11722 fftfilt=dc_Y=128:weight_Y='squish(1-(Y+X)/100)'
11724 • Low-pass:
11726 fftfilt=dc_Y=0:weight_Y='squish((Y+X)/100-1)'
11728 • Sharpen:
11730 fftfilt=dc_Y=0:weight_Y='1+squish(1-(Y+X)/100)'
11732 • Blur:
11734 fftfilt=dc_Y=0:weight_Y='exp(-4 * ((Y+X)/(W+H)))'
11736 field
11738 Extract a single field from an interlaced image using stride arithmetic
11739 to avoid wasting CPU time. The output frames are marked as non-
11740 interlaced.
11741 The filter accepts the following options:
11743 type
11745 Specify whether to extract the top (if the value is 0 or "top") or
11746 the bottom field (if the value is 1 or "bottom").
11747 fieldhint
11749 Create new frames by copying the top and bottom fields from surrounding
11750 frames supplied as numbers by the hint file.
11751 hint
11753 Set file containing hints: absolute/relative frame numbers.
11754 There must be one line for each frame in a clip. Each line must
11756 contain two numbers separated by the comma, optionally followed by
11757 "-" or "+". Numbers supplied on each line of file can not be out
11758 of [N-1,N+1] where N is current frame number for "absolute" mode or
11759 out of [-1, 1] range for "relative" mode. First number tells from
11760 which frame to pick up top field and second number tells from which
11761 frame to pick up bottom field.
11762 If optionally followed by "+" output frame will be marked as
11764 interlaced, else if followed by "-" output frame will be marked as
11765 progressive, else it will be marked same as input frame. If
11766 optionally followed by "t" output frame will use only top field, or
11767 in case of "b" it will use only bottom field. If line starts with
11768 "#" or ";" that line is skipped.
11769 mode
11771 Can be item "absolute" or "relative" or "pattern". Default is
11772 "absolute". The "pattern" mode is same as "relative" mode, except
11773 at last entry of file if there are more frames to process than
11774 "hint" file is seek back to start.
11775 Example of first several lines of "hint" file for "relative" mode:
11777 0,0 - # first frame
11779 1,0 - # second frame, use third's frame top field and second's frame bottom field
11780 1,0 - # third frame, use fourth's frame top field and third's frame bottom field
11781 1,0 -
11782 0,0 -
11783 0,0 -
11784 1,0 -
11785 1,0 -
11786 1,0 -
11787 0,0 -
11788 0,0 -
11789 1,0 -
11790 1,0 -
11791 1,0 -
11792 0,0 -
11793 fieldmatch
11795 Field matching filter for inverse telecine. It is meant to reconstruct
11796 the progressive frames from a telecined stream. The filter does not
11797 drop duplicated frames, so to achieve a complete inverse telecine
11798 "fieldmatch" needs to be followed by a decimation filter such as
11799 decimate in the filtergraph.
11800 The separation of the field matching and the decimation is notably
11802 motivated by the possibility of inserting a de-interlacing filter
11803 fallback between the two. If the source has mixed telecined and real
11804 interlaced content, "fieldmatch" will not be able to match fields for
11805 the interlaced parts. But these remaining combed frames will be marked
11806 as interlaced, and thus can be de-interlaced by a later filter such as
11807 yadif before decimation.
11808 In addition to the various configuration options, "fieldmatch" can take
11810 an optional second stream, activated through the ppsrc option. If
11811 enabled, the frames reconstruction will be based on the fields and
11812 frames from this second stream. This allows the first input to be pre-
11813 processed in order to help the various algorithms of the filter, while
11814 keeping the output lossless (assuming the fields are matched properly).
11815 Typically, a field-aware denoiser, or brightness/contrast adjustments
11816 can help.
11817 Note that this filter uses the same algorithms as TIVTC/TFM (AviSynth
11819 project) and VIVTC/VFM (VapourSynth project). The later is a light
11820 clone of TFM from which "fieldmatch" is based on. While the semantic
11821 and usage are very close, some behaviour and options names can differ.
11822 The decimate filter currently only works for constant frame rate input.
11824 If your input has mixed telecined (30fps) and progressive content with
11825 a lower framerate like 24fps use the following filterchain to produce
11826 the necessary cfr stream:
11827 "dejudder,fps=30000/1001,fieldmatch,decimate".
11828 The filter accepts the following options:
11830 order
11832 Specify the assumed field order of the input stream. Available
11833 values are:
11834 auto
11836 Auto detect parity (use FFmpeg's internal parity value).
11837 bff Assume bottom field first.
11839 tff Assume top field first.
11841 Note that it is sometimes recommended not to trust the parity
11843 announced by the stream.
11844 Default value is auto.
11846 mode
11848 Set the matching mode or strategy to use. pc mode is the safest in
11849 the sense that it won't risk creating jerkiness due to duplicate
11850 frames when possible, but if there are bad edits or blended fields
11851 it will end up outputting combed frames when a good match might
11852 actually exist. On the other hand, pcn_ub mode is the most risky in
11853 terms of creating jerkiness, but will almost always find a good
11854 frame if there is one. The other values are all somewhere in
11855 between pc and pcn_ub in terms of risking jerkiness and creating
11856 duplicate frames versus finding good matches in sections with bad
11857 edits, orphaned fields, blended fields, etc.
11858 More details about p/c/n/u/b are available in p/c/n/u/b meaning
11860 section.
11861 Available values are:
11863 pc 2-way matching (p/c)
11865 pc_n
11867 2-way matching, and trying 3rd match if still combed (p/c + n)
11868 pc_u
11870 2-way matching, and trying 3rd match (same order) if still
11871 combed (p/c + u)
11872 pc_n_ub
11874 2-way matching, trying 3rd match if still combed, and trying
11875 4th/5th matches if still combed (p/c + n + u/b)
11876 pcn 3-way matching (p/c/n)
11878 pcn_ub
11880 3-way matching, and trying 4th/5th matches if all 3 of the
11881 original matches are detected as combed (p/c/n + u/b)
11882 The parenthesis at the end indicate the matches that would be used
11884 for that mode assuming order=tff (and field on auto or top).
11885 In terms of speed pc mode is by far the fastest and pcn_ub is the
11887 slowest.
11888 Default value is pc_n.
11890 ppsrc
11892 Mark the main input stream as a pre-processed input, and enable the
11893 secondary input stream as the clean source to pick the fields from.
11894 See the filter introduction for more details. It is similar to the
11895 clip2 feature from VFM/TFM.
11896 Default value is 0 (disabled).
11898 field
11900 Set the field to match from. It is recommended to set this to the
11901 same value as order unless you experience matching failures with
11902 that setting. In certain circumstances changing the field that is
11903 used to match from can have a large impact on matching performance.
11904 Available values are:
11905 auto
11907 Automatic (same value as order).
11908 bottom
11910 Match from the bottom field.
11911 top Match from the top field.
11913 Default value is auto.
11915 mchroma
11917 Set whether or not chroma is included during the match comparisons.
11918 In most cases it is recommended to leave this enabled. You should
11919 set this to 0 only if your clip has bad chroma problems such as
11920 heavy rainbowing or other artifacts. Setting this to 0 could also
11921 be used to speed things up at the cost of some accuracy.
11922 Default value is 1.
11924 y0
11926 y1 These define an exclusion band which excludes the lines between y0
11927 and y1 from being included in the field matching decision. An
11928 exclusion band can be used to ignore subtitles, a logo, or other
11929 things that may interfere with the matching. y0 sets the starting
11930 scan line and y1 sets the ending line; all lines in between y0 and
11931 y1 (including y0 and y1) will be ignored. Setting y0 and y1 to the
11932 same value will disable the feature. y0 and y1 defaults to 0.
11933 scthresh
11935 Set the scene change detection threshold as a percentage of maximum
11936 change on the luma plane. Good values are in the "[8.0, 14.0]"
11937 range. Scene change detection is only relevant in case
11938 combmatch=sc. The range for scthresh is "[0.0, 100.0]".
11939 Default value is 12.0.
11941 combmatch
11943 When combatch is not none, "fieldmatch" will take into account the
11944 combed scores of matches when deciding what match to use as the
11945 final match. Available values are:
11946 none
11948 No final matching based on combed scores.
11949 sc Combed scores are only used when a scene change is detected.
11951 full
11953 Use combed scores all the time.
11954 Default is sc.
11956 combdbg
11958 Force "fieldmatch" to calculate the combed metrics for certain
11959 matches and print them. This setting is known as micout in TFM/VFM
11960 vocabulary. Available values are:
11961 none
11963 No forced calculation.
11964 pcn Force p/c/n calculations.
11966 pcnub
11968 Force p/c/n/u/b calculations.
11969 Default value is none.
11971 cthresh
11973 This is the area combing threshold used for combed frame detection.
11974 This essentially controls how "strong" or "visible" combing must be
11975 to be detected. Larger values mean combing must be more visible
11976 and smaller values mean combing can be less visible or strong and
11977 still be detected. Valid settings are from -1 (every pixel will be
11978 detected as combed) to 255 (no pixel will be detected as combed).
11979 This is basically a pixel difference value. A good range is "[8,
11980 12]".
11981 Default value is 9.
11983 chroma
11985 Sets whether or not chroma is considered in the combed frame
11986 decision. Only disable this if your source has chroma problems
11987 (rainbowing, etc.) that are causing problems for the combed frame
11988 detection with chroma enabled. Actually, using chroma=0 is usually
11989 more reliable, except for the case where there is chroma only
11990 combing in the source.
11991 Default value is 0.
11993 blockx
11995 blocky
11996 Respectively set the x-axis and y-axis size of the window used
11997 during combed frame detection. This has to do with the size of the
11998 area in which combpel pixels are required to be detected as combed
11999 for a frame to be declared combed. See the combpel parameter
12000 description for more info. Possible values are any number that is
12001 a power of 2 starting at 4 and going up to 512.
12002 Default value is 16.
12004 combpel
12006 The number of combed pixels inside any of the blocky by blockx size
12007 blocks on the frame for the frame to be detected as combed. While
12008 cthresh controls how "visible" the combing must be, this setting
12009 controls "how much" combing there must be in any localized area (a
12010 window defined by the blockx and blocky settings) on the frame.
12011 Minimum value is 0 and maximum is "blocky x blockx" (at which point
12012 no frames will ever be detected as combed). This setting is known
12013 as MI in TFM/VFM vocabulary.
12014 Default value is 80.
12016 p/c/n/u/b meaning
12018 p/c/n
12020 We assume the following telecined stream:
12022 Top fields: 1 2 2 3 4
12024 Bottom fields: 1 2 3 4 4
12025 The numbers correspond to the progressive frame the fields relate to.
12027 Here, the first two frames are progressive, the 3rd and 4th are combed,
12028 and so on.
12029 When "fieldmatch" is configured to run a matching from bottom
12031 (field=bottom) this is how this input stream get transformed:
12032 Input stream:
12034 T 1 2 2 3 4
12035 B 1 2 3 4 4 <-- matching reference
12036 Matches: c c n n c
12038 Output stream:
12040 T 1 2 3 4 4
12041 B 1 2 3 4 4
12042 As a result of the field matching, we can see that some frames get
12044 duplicated. To perform a complete inverse telecine, you need to rely
12045 on a decimation filter after this operation. See for instance the
12046 decimate filter.
12047 The same operation now matching from top fields (field=top) looks like
12049 this:
12050 Input stream:
12052 T 1 2 2 3 4 <-- matching reference
12053 B 1 2 3 4 4
12054 Matches: c c p p c
12056 Output stream:
12058 T 1 2 2 3 4
12059 B 1 2 2 3 4
12060 In these examples, we can see what p, c and n mean; basically, they
12062 refer to the frame and field of the opposite parity:
12063 *<p matches the field of the opposite parity in the previous frame>
12065 *<c matches the field of the opposite parity in the current frame>
12066 *<n matches the field of the opposite parity in the next frame>
12067 u/b
12069 The u and b matching are a bit special in the sense that they match
12071 from the opposite parity flag. In the following examples, we assume
12072 that we are currently matching the 2nd frame (Top:2, bottom:2).
12073 According to the match, a 'x' is placed above and below each matched
12074 fields.
12075 With bottom matching (field=bottom):
12077 Match: c p n b u
12079 x x x x x
12081 Top 1 2 2 1 2 2 1 2 2 1 2 2 1 2 2
12082 Bottom 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3
12083 x x x x x
12084 Output frames:
12086 2 1 2 2 2
12087 2 2 2 1 3
12088 With top matching (field=top):
12090 Match: c p n b u
12092 x x x x x
12094 Top 1 2 2 1 2 2 1 2 2 1 2 2 1 2 2
12095 Bottom 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3
12096 x x x x x
12097 Output frames:
12099 2 2 2 1 2
12100 2 1 3 2 2
12101 Examples
12103 Simple IVTC of a top field first telecined stream:
12105 fieldmatch=order=tff:combmatch=none, decimate
12107 Advanced IVTC, with fallback on yadif for still combed frames:
12109 fieldmatch=order=tff:combmatch=full, yadif=deint=interlaced, decimate
12111 fieldorder
12113 Transform the field order of the input video.
12114 It accepts the following parameters:
12116 order
12118 The output field order. Valid values are tff for top field first or
12119 bff for bottom field first.
12120 The default value is tff.
12122 The transformation is done by shifting the picture content up or down
12124 by one line, and filling the remaining line with appropriate picture
12125 content. This method is consistent with most broadcast field order
12126 converters.
12127 If the input video is not flagged as being interlaced, or it is already
12129 flagged as being of the required output field order, then this filter
12130 does not alter the incoming video.
12131 It is very useful when converting to or from PAL DV material, which is
12133 bottom field first.
12134 For example:
12136 ffmpeg -i in.vob -vf "fieldorder=bff" out.dv
12138 fifo, afifo
12140 Buffer input images and send them when they are requested.
12141 It is mainly useful when auto-inserted by the libavfilter framework.
12143 It does not take parameters.
12145 fillborders
12147 Fill borders of the input video, without changing video stream
12148 dimensions. Sometimes video can have garbage at the four edges and you
12149 may not want to crop video input to keep size multiple of some number.
12150 This filter accepts the following options:
12152 left
12154 Number of pixels to fill from left border.
12155 right
12157 Number of pixels to fill from right border.
12158 top Number of pixels to fill from top border.
12160 bottom
12162 Number of pixels to fill from bottom border.
12163 mode
12165 Set fill mode.
12166 It accepts the following values:
12168 smear
12170 fill pixels using outermost pixels
12171 mirror
12173 fill pixels using mirroring (half sample symmetric)
12174 fixed
12176 fill pixels with constant value
12177 reflect
12179 fill pixels using reflecting (whole sample symmetric)
12180 wrap
12182 fill pixels using wrapping
12183 fade
12185 fade pixels to constant value
12186 margins
12188 fill pixels at top and bottom with weighted averages pixels
12189 near borders
12190 Default is smear.
12192 color
12194 Set color for pixels in fixed or fade mode. Default is black.
12195 Commands
12197 This filter supports same commands as options. The command accepts the
12199 same syntax of the corresponding option.
12200 If the specified expression is not valid, it is kept at its current
12202 value.
12203 find_rect
12205 Find a rectangular object
12206 It accepts the following options:
12208 object
12210 Filepath of the object image, needs to be in gray8.
12211 threshold
12213 Detection threshold, default is 0.5.
12214 mipmaps
12216 Number of mipmaps, default is 3.
12217 xmin, ymin, xmax, ymax
12219 Specifies the rectangle in which to search.
12220 discard
12222 Discard frames where object is not detected. Default is disabled.
12223 Examples
12225 • Cover a rectangular object by the supplied image of a given video
12227 using ffmpeg:
12228 ffmpeg -i file.ts -vf find_rect=newref.pgm,cover_rect=cover.jpg:mode=cover new.mkv
12230 floodfill
12232 Flood area with values of same pixel components with another values.
12233 It accepts the following options:
12235 x Set pixel x coordinate.
12237 y Set pixel y coordinate.
12239 s0 Set source #0 component value.
12241 s1 Set source #1 component value.
12243 s2 Set source #2 component value.
12245 s3 Set source #3 component value.
12247 d0 Set destination #0 component value.
12249 d1 Set destination #1 component value.
12251 d2 Set destination #2 component value.
12253 d3 Set destination #3 component value.
12255 format
12257 Convert the input video to one of the specified pixel formats.
12258 Libavfilter will try to pick one that is suitable as input to the next
12259 filter.
12260 It accepts the following parameters:
12262 pix_fmts
12264 A '|'-separated list of pixel format names, such as
12265 "pix_fmts=yuv420p|monow|rgb24".
12266 Examples
12268 • Convert the input video to the yuv420p format
12270 format=pix_fmts=yuv420p
12272 Convert the input video to any of the formats in the list
12274 format=pix_fmts=yuv420p|yuv444p|yuv410p
12276 fps
12278 Convert the video to specified constant frame rate by duplicating or
12279 dropping frames as necessary.
12280 It accepts the following parameters:
12282 fps The desired output frame rate. It accepts expressions containing
12284 the following constants:
12285 source_fps
12287 The input's frame rate
12288 ntsc
12290 NTSC frame rate of "30000/1001"
12291 pal PAL frame rate of 25.0
12293 film
12295 Film frame rate of 24.0
12296 ntsc_film
12298 NTSC-film frame rate of "24000/1001"
12299 The default is 25.
12301 start_time
12303 Assume the first PTS should be the given value, in seconds. This
12304 allows for padding/trimming at the start of stream. By default, no
12305 assumption is made about the first frame's expected PTS, so no
12306 padding or trimming is done. For example, this could be set to 0
12307 to pad the beginning with duplicates of the first frame if a video
12308 stream starts after the audio stream or to trim any frames with a
12309 negative PTS.
12310 round
12312 Timestamp (PTS) rounding method.
12313 Possible values are:
12315 zero
12317 round towards 0
12318 inf round away from 0
12320 down
12322 round towards -infinity
12323 up round towards +infinity
12325 near
12327 round to nearest
12328 The default is "near".
12330 eof_action
12332 Action performed when reading the last frame.
12333 Possible values are:
12335 round
12337 Use same timestamp rounding method as used for other frames.
12338 pass
12340 Pass through last frame if input duration has not been reached
12341 yet.
12342 The default is "round".
12344 Alternatively, the options can be specified as a flat string:
12346 fps[:start_time[:round]].
12347 See also the setpts filter.
12349 Examples
12351 • A typical usage in order to set the fps to 25:
12353 fps=fps=25
12355 • Sets the fps to 24, using abbreviation and rounding method to round
12357 to nearest:
12358 fps=fps=film:round=near
12360 framepack
12362 Pack two different video streams into a stereoscopic video, setting
12363 proper metadata on supported codecs. The two views should have the same
12364 size and framerate and processing will stop when the shorter video
12365 ends. Please note that you may conveniently adjust view properties with
12366 the scale and fps filters.
12367 It accepts the following parameters:
12369 format
12371 The desired packing format. Supported values are:
12372 sbs The views are next to each other (default).
12374 tab The views are on top of each other.
12376 lines
12378 The views are packed by line.
12379 columns
12381 The views are packed by column.
12382 frameseq
12384 The views are temporally interleaved.
12385 Some examples:
12387 # Convert left and right views into a frame-sequential video
12389 ffmpeg -i LEFT -i RIGHT -filter_complex framepack=frameseq OUTPUT
12390 # Convert views into a side-by-side video with the same output resolution as the input
12392 ffmpeg -i LEFT -i RIGHT -filter_complex [0:v]scale=w=iw/2[left],[1:v]scale=w=iw/2[right],[left][right]framepack=sbs OUTPUT
12393 framerate
12395 Change the frame rate by interpolating new video output frames from the
12396 source frames.
12397 This filter is not designed to function correctly with interlaced
12399 media. If you wish to change the frame rate of interlaced media then
12400 you are required to deinterlace before this filter and re-interlace
12401 after this filter.
12402 A description of the accepted options follows.
12404 fps Specify the output frames per second. This option can also be
12406 specified as a value alone. The default is 50.
12407 interp_start
12409 Specify the start of a range where the output frame will be created
12410 as a linear interpolation of two frames. The range is [0-255], the
12411 default is 15.
12412 interp_end
12414 Specify the end of a range where the output frame will be created
12415 as a linear interpolation of two frames. The range is [0-255], the
12416 default is 240.
12417 scene
12419 Specify the level at which a scene change is detected as a value
12420 between 0 and 100 to indicate a new scene; a low value reflects a
12421 low probability for the current frame to introduce a new scene,
12422 while a higher value means the current frame is more likely to be
12423 one. The default is 8.2.
12424 flags
12426 Specify flags influencing the filter process.
12427 Available value for flags is:
12429 scene_change_detect, scd
12431 Enable scene change detection using the value of the option
12432 scene. This flag is enabled by default.
12433 framestep
12435 Select one frame every N-th frame.
12436 This filter accepts the following option:
12438 step
12440 Select frame after every "step" frames. Allowed values are
12441 positive integers higher than 0. Default value is 1.
12442 freezedetect
12444 Detect frozen video.
12445 This filter logs a message and sets frame metadata when it detects that
12447 the input video has no significant change in content during a specified
12448 duration. Video freeze detection calculates the mean average absolute
12449 difference of all the components of video frames and compares it to a
12450 noise floor.
12451 The printed times and duration are expressed in seconds. The
12453 "lavfi.freezedetect.freeze_start" metadata key is set on the first
12454 frame whose timestamp equals or exceeds the detection duration and it
12455 contains the timestamp of the first frame of the freeze. The
12456 "lavfi.freezedetect.freeze_duration" and
12457 "lavfi.freezedetect.freeze_end" metadata keys are set on the first
12458 frame after the freeze.
12459 The filter accepts the following options:
12461 noise, n
12463 Set noise tolerance. Can be specified in dB (in case "dB" is
12464 appended to the specified value) or as a difference ratio between 0
12465 and 1. Default is -60dB, or 0.001.
12466 duration, d
12468 Set freeze duration until notification (default is 2 seconds).
12469 freezeframes
12471 Freeze video frames.
12472 This filter freezes video frames using frame from 2nd input.
12474 The filter accepts the following options:
12476 first
12478 Set number of first frame from which to start freeze.
12479 last
12481 Set number of last frame from which to end freeze.
12482 replace
12484 Set number of frame from 2nd input which will be used instead of
12485 replaced frames.
12486 frei0r
12488 Apply a frei0r effect to the input video.
12489 To enable the compilation of this filter, you need to install the
12491 frei0r header and configure FFmpeg with "--enable-frei0r".
12492 It accepts the following parameters:
12494 filter_name
12496 The name of the frei0r effect to load. If the environment variable
12497 FREI0R_PATH is defined, the frei0r effect is searched for in each
12498 of the directories specified by the colon-separated list in
12499 FREI0R_PATH. Otherwise, the standard frei0r paths are searched, in
12500 this order: HOME/.frei0r-1/lib/, /usr/local/lib/frei0r-1/,
12501 /usr/lib/frei0r-1/.
12502 filter_params
12504 A '|'-separated list of parameters to pass to the frei0r effect.
12505 A frei0r effect parameter can be a boolean (its value is either "y" or
12507 "n"), a double, a color (specified as R/G/B, where R, G, and B are
12508 floating point numbers between 0.0 and 1.0, inclusive) or a color
12509 description as specified in the "Color" section in the ffmpeg-utils
12510 manual, a position (specified as X/Y, where X and Y are floating point
12511 numbers) and/or a string.
12512 The number and types of parameters depend on the loaded effect. If an
12514 effect parameter is not specified, the default value is set.
12515 Examples
12517 • Apply the distort0r effect, setting the first two double
12519 parameters:
12520 frei0r=filter_name=distort0r:filter_params=0.5|0.01
12522 • Apply the colordistance effect, taking a color as the first
12524 parameter:
12525 frei0r=colordistance:0.2/0.3/0.4
12527 frei0r=colordistance:violet
12528 frei0r=colordistance:0x112233
12529 • Apply the perspective effect, specifying the top left and top right
12531 image positions:
12532 frei0r=perspective:0.2/0.2|0.8/0.2
12534 For more information, see <http://frei0r.dyne.org>
12536 Commands
12538 This filter supports the filter_params option as commands.
12540 fspp
12542 Apply fast and simple postprocessing. It is a faster version of spp.
12543 It splits (I)DCT into horizontal/vertical passes. Unlike the simple
12545 post- processing filter, one of them is performed once per block, not
12546 per pixel. This allows for much higher speed.
12547 The filter accepts the following options:
12549 quality
12551 Set quality. This option defines the number of levels for
12552 averaging. It accepts an integer in the range 4-5. Default value is
12553 4.
12554 qp Force a constant quantization parameter. It accepts an integer in
12556 range 0-63. If not set, the filter will use the QP from the video
12557 stream (if available).
12558 strength
12560 Set filter strength. It accepts an integer in range -15 to 32.
12561 Lower values mean more details but also more artifacts, while
12562 higher values make the image smoother but also blurrier. Default
12563 value is 0 − PSNR optimal.
12564 use_bframe_qp
12566 Enable the use of the QP from the B-Frames if set to 1. Using this
12567 option may cause flicker since the B-Frames have often larger QP.
12568 Default is 0 (not enabled).
12569 gblur
12571 Apply Gaussian blur filter.
12572 The filter accepts the following options:
12574 sigma
12576 Set horizontal sigma, standard deviation of Gaussian blur. Default
12577 is 0.5.
12578 steps
12580 Set number of steps for Gaussian approximation. Default is 1.
12581 planes
12583 Set which planes to filter. By default all planes are filtered.
12584 sigmaV
12586 Set vertical sigma, if negative it will be same as "sigma".
12587 Default is -1.
12588 Commands
12590 This filter supports same commands as options. The command accepts the
12592 same syntax of the corresponding option.
12593 If the specified expression is not valid, it is kept at its current
12595 value.
12596 geq
12598 Apply generic equation to each pixel.
12599 The filter accepts the following options:
12601 lum_expr, lum
12603 Set the luminance expression.
12604 cb_expr, cb
12606 Set the chrominance blue expression.
12607 cr_expr, cr
12609 Set the chrominance red expression.
12610 alpha_expr, a
12612 Set the alpha expression.
12613 red_expr, r
12615 Set the red expression.
12616 green_expr, g
12618 Set the green expression.
12619 blue_expr, b
12621 Set the blue expression.
12622 The colorspace is selected according to the specified options. If one
12624 of the lum_expr, cb_expr, or cr_expr options is specified, the filter
12625 will automatically select a YCbCr colorspace. If one of the red_expr,
12626 green_expr, or blue_expr options is specified, it will select an RGB
12627 colorspace.
12628 If one of the chrominance expression is not defined, it falls back on
12630 the other one. If no alpha expression is specified it will evaluate to
12631 opaque value. If none of chrominance expressions are specified, they
12632 will evaluate to the luminance expression.
12633 The expressions can use the following variables and functions:
12635 N The sequential number of the filtered frame, starting from 0.
12637 X
12639 Y The coordinates of the current sample.
12640 W
12642 H The width and height of the image.
12643 SW
12645 SH Width and height scale depending on the currently filtered plane.
12646 It is the ratio between the corresponding luma plane number of
12647 pixels and the current plane ones. E.g. for YUV4:2:0 the values are
12648 "1,1" for the luma plane, and "0.5,0.5" for chroma planes.
12649 T Time of the current frame, expressed in seconds.
12651 p(x, y)
12653 Return the value of the pixel at location (x,y) of the current
12654 plane.
12655 lum(x, y)
12657 Return the value of the pixel at location (x,y) of the luminance
12658 plane.
12659 cb(x, y)
12661 Return the value of the pixel at location (x,y) of the blue-
12662 difference chroma plane. Return 0 if there is no such plane.
12663 cr(x, y)
12665 Return the value of the pixel at location (x,y) of the red-
12666 difference chroma plane. Return 0 if there is no such plane.
12667 r(x, y)
12669 g(x, y)
12670 b(x, y)
12671 Return the value of the pixel at location (x,y) of the
12672 red/green/blue component. Return 0 if there is no such component.
12673 alpha(x, y)
12675 Return the value of the pixel at location (x,y) of the alpha plane.
12676 Return 0 if there is no such plane.
12677 psum(x,y), lumsum(x, y), cbsum(x,y), crsum(x,y), rsum(x,y), gsum(x,y),
12679 bsum(x,y), alphasum(x,y)
12680 Sum of sample values in the rectangle from (0,0) to (x,y), this
12681 allows obtaining sums of samples within a rectangle. See the
12682 functions without the sum postfix.
12683 interpolation
12685 Set one of interpolation methods:
12686 nearest, n
12688 bilinear, b
12689 Default is bilinear.
12691 For functions, if x and y are outside the area, the value will be
12693 automatically clipped to the closer edge.
12694 Please note that this filter can use multiple threads in which case
12696 each slice will have its own expression state. If you want to use only
12697 a single expression state because your expressions depend on previous
12698 state then you should limit the number of filter threads to 1.
12699 Examples
12701 • Flip the image horizontally:
12703 geq=p(W-X\,Y)
12705 • Generate a bidimensional sine wave, with angle "PI/3" and a
12707 wavelength of 100 pixels:
12708 geq=128 + 100*sin(2*(PI/100)*(cos(PI/3)*(X-50*T) + sin(PI/3)*Y)):128:128
12710 • Generate a fancy enigmatic moving light:
12712 nullsrc=s=256x256,geq=random(1)/hypot(X-cos(N*0.07)*W/2-W/2\,Y-sin(N*0.09)*H/2-H/2)^2*1000000*sin(N*0.02):128:128
12714 • Generate a quick emboss effect:
12716 format=gray,geq=lum_expr='(p(X,Y)+(256-p(X-4,Y-4)))/2'
12718 • Modify RGB components depending on pixel position:
12720 geq=r='X/W*r(X,Y)':g='(1-X/W)*g(X,Y)':b='(H-Y)/H*b(X,Y)'
12722 • Create a radial gradient that is the same size as the input (also
12724 see the vignette filter):
12725 geq=lum=255*gauss((X/W-0.5)*3)*gauss((Y/H-0.5)*3)/gauss(0)/gauss(0),format=gray
12727 gradfun
12729 Fix the banding artifacts that are sometimes introduced into nearly
12730 flat regions by truncation to 8-bit color depth. Interpolate the
12731 gradients that should go where the bands are, and dither them.
12732 It is designed for playback only. Do not use it prior to lossy
12734 compression, because compression tends to lose the dither and bring
12735 back the bands.
12736 It accepts the following parameters:
12738 strength
12740 The maximum amount by which the filter will change any one pixel.
12741 This is also the threshold for detecting nearly flat regions.
12742 Acceptable values range from .51 to 64; the default value is 1.2.
12743 Out-of-range values will be clipped to the valid range.
12744 radius
12746 The neighborhood to fit the gradient to. A larger radius makes for
12747 smoother gradients, but also prevents the filter from modifying the
12748 pixels near detailed regions. Acceptable values are 8-32; the
12749 default value is 16. Out-of-range values will be clipped to the
12750 valid range.
12751 Alternatively, the options can be specified as a flat string:
12753 strength[:radius]
12754 Examples
12756 • Apply the filter with a 3.5 strength and radius of 8:
12758 gradfun=3.5:8
12760 • Specify radius, omitting the strength (which will fall-back to the
12762 default value):
12763 gradfun=radius=8
12765 graphmonitor
12767 Show various filtergraph stats.
12768 With this filter one can debug complete filtergraph. Especially issues
12770 with links filling with queued frames.
12771 The filter accepts the following options:
12773 size, s
12775 Set video output size. Default is hd720.
12776 opacity, o
12778 Set video opacity. Default is 0.9. Allowed range is from 0 to 1.
12779 mode, m
12781 Set output mode, can be fulll or compact. In compact mode only
12782 filters with some queued frames have displayed stats.
12783 flags, f
12785 Set flags which enable which stats are shown in video.
12786 Available values for flags are:
12788 queue
12790 Display number of queued frames in each link.
12791 frame_count_in
12793 Display number of frames taken from filter.
12794 frame_count_out
12796 Display number of frames given out from filter.
12797 frame_count_delta
12799 Display delta number of frames between above two values.
12800 pts Display current filtered frame pts.
12802 pts_delta
12804 Display pts delta between current and previous frame.
12805 time
12807 Display current filtered frame time.
12808 time_delta
12810 Display time delta between current and previous frame.
12811 timebase
12813 Display time base for filter link.
12814 format
12816 Display used format for filter link.
12817 size
12819 Display video size or number of audio channels in case of audio
12820 used by filter link.
12821 rate
12823 Display video frame rate or sample rate in case of audio used
12824 by filter link.
12825 eof Display link output status.
12827 sample_count_in
12829 Display number of samples taken from filter.
12830 sample_count_out
12832 Display number of samples given out from filter.
12833 sample_count_delta
12835 Display delta number of samples between above two values.
12836 rate, r
12838 Set upper limit for video rate of output stream, Default value is
12839 25. This guarantee that output video frame rate will not be higher
12840 than this value.
12841 grayworld
12843 A color constancy filter that applies color correction based on the
12844 grayworld assumption
12845 See:
12847 <https://www.researchgate.net/publication/275213614_A_New_Color_Correction_Method_for_Underwater_Imaging>
12848 The algorithm uses linear light, so input data should be linearized
12850 beforehand (and possibly correctly tagged).
12851 ffmpeg -i INPUT -vf zscale=transfer=linear,grayworld,zscale=transfer=bt709,format=yuv420p OUTPUT
12853 greyedge
12855 A color constancy variation filter which estimates scene illumination
12856 via grey edge algorithm and corrects the scene colors accordingly.
12857 See: <https://staff.science.uva.nl/th.gevers/pub/GeversTIP07.pdf>
12859 The filter accepts the following options:
12861 difford
12863 The order of differentiation to be applied on the scene. Must be
12864 chosen in the range [0,2] and default value is 1.
12865 minknorm
12867 The Minkowski parameter to be used for calculating the Minkowski
12868 distance. Must be chosen in the range [0,20] and default value is
12869 1. Set to 0 for getting max value instead of calculating Minkowski
12870 distance.
12871 sigma
12873 The standard deviation of Gaussian blur to be applied on the scene.
12874 Must be chosen in the range [0,1024.0] and default value = 1.
12875 floor( sigma * break_off_sigma(3) ) can't be equal to 0 if difford
12876 is greater than 0.
12877 Examples
12879 • Grey Edge:
12881 greyedge=difford=1:minknorm=5:sigma=2
12883 • Max Edge:
12885 greyedge=difford=1:minknorm=0:sigma=2
12887 guided
12889 Apply guided filter for edge-preserving smoothing, dehazing and so on.
12890 The filter accepts the following options:
12892 radius
12894 Set the box radius in pixels. Allowed range is 1 to 20. Default is
12895 3.
12896 eps Set regularization parameter (with square). Allowed range is 0 to
12898 1. Default is 0.01.
12899 mode
12901 Set filter mode. Can be "basic" or "fast". Default is "basic".
12902 sub Set subsampling ratio for "fast" mode. Range is 2 to 64. Default
12904 is 4. No subsampling occurs in "basic" mode.
12905 guidance
12907 Set guidance mode. Can be "off" or "on". Default is "off". If
12908 "off", single input is required. If "on", two inputs of the same
12909 resolution and pixel format are required. The second input serves
12910 as the guidance.
12911 planes
12913 Set planes to filter. Default is first only.
12914 Commands
12916 This filter supports the all above options as commands.
12918 Examples
12920 • Edge-preserving smoothing with guided filter:
12922 ffmpeg -i in.png -vf guided out.png
12924 • Dehazing, structure-transferring filtering, detail enhancement with
12926 guided filter. For the generation of guidance image, refer to
12927 paper "Guided Image Filtering". See:
12928 <http://kaiminghe.com/publications/pami12guidedfilter.pdf>.
12929 ffmpeg -i in.png -i guidance.png -filter_complex guided=guidance=on out.png
12931 haldclut
12933 Apply a Hald CLUT to a video stream.
12934 First input is the video stream to process, and second one is the Hald
12936 CLUT. The Hald CLUT input can be a simple picture or a complete video
12937 stream.
12938 The filter accepts the following options:
12940 clut
12942 Set which CLUT video frames will be processed from second input
12943 stream, can be first or all. Default is all.
12944 shortest
12946 Force termination when the shortest input terminates. Default is 0.
12947 repeatlast
12949 Continue applying the last CLUT after the end of the stream. A
12950 value of 0 disable the filter after the last frame of the CLUT is
12951 reached. Default is 1.
12952 "haldclut" also has the same interpolation options as lut3d (both
12954 filters share the same internals).
12955 This filter also supports the framesync options.
12957 More information about the Hald CLUT can be found on Eskil Steenberg's
12959 website (Hald CLUT author) at
12960 <http://www.quelsolaar.com/technology/clut.html>.
12961 Commands
12963 This filter supports the "interp" option as commands.
12965 Workflow examples
12967 Hald CLUT video stream
12969 Generate an identity Hald CLUT stream altered with various effects:
12971 ffmpeg -f lavfi -i B<haldclutsrc>=8 -vf "hue=H=2*PI*t:s=sin(2*PI*t)+1, curves=cross_process" -t 10 -c:v ffv1 clut.nut
12973 Note: make sure you use a lossless codec.
12975 Then use it with "haldclut" to apply it on some random stream:
12977 ffmpeg -f lavfi -i mandelbrot -i clut.nut -filter_complex '[0][1] haldclut' -t 20 mandelclut.mkv
12979 The Hald CLUT will be applied to the 10 first seconds (duration of
12981 clut.nut), then the latest picture of that CLUT stream will be applied
12982 to the remaining frames of the "mandelbrot" stream.
12983 Hald CLUT with preview
12985 A Hald CLUT is supposed to be a squared image of "Level*Level*Level" by
12987 "Level*Level*Level" pixels. For a given Hald CLUT, FFmpeg will select
12988 the biggest possible square starting at the top left of the picture.
12989 The remaining padding pixels (bottom or right) will be ignored. This
12990 area can be used to add a preview of the Hald CLUT.
12991 Typically, the following generated Hald CLUT will be supported by the
12993 "haldclut" filter:
12994 ffmpeg -f lavfi -i B<haldclutsrc>=8 -vf "
12996 pad=iw+320 [padded_clut];
12997 smptebars=s=320x256, split [a][b];
12998 [padded_clut][a] overlay=W-320:h, curves=color_negative [main];
12999 [main][b] overlay=W-320" -frames:v 1 clut.png
13000 It contains the original and a preview of the effect of the CLUT: SMPTE
13002 color bars are displayed on the right-top, and below the same color
13003 bars processed by the color changes.
13004 Then, the effect of this Hald CLUT can be visualized with:
13006 ffplay input.mkv -vf "movie=clut.png, [in] haldclut"
13008 hflip
13010 Flip the input video horizontally.
13011 For example, to horizontally flip the input video with ffmpeg:
13013 ffmpeg -i in.avi -vf "hflip" out.avi
13015 histeq
13017 This filter applies a global color histogram equalization on a per-
13018 frame basis.
13019 It can be used to correct video that has a compressed range of pixel
13021 intensities. The filter redistributes the pixel intensities to
13022 equalize their distribution across the intensity range. It may be
13023 viewed as an "automatically adjusting contrast filter". This filter is
13024 useful only for correcting degraded or poorly captured source video.
13025 The filter accepts the following options:
13027 strength
13029 Determine the amount of equalization to be applied. As the
13030 strength is reduced, the distribution of pixel intensities more-
13031 and-more approaches that of the input frame. The value must be a
13032 float number in the range [0,1] and defaults to 0.200.
13033 intensity
13035 Set the maximum intensity that can generated and scale the output
13036 values appropriately. The strength should be set as desired and
13037 then the intensity can be limited if needed to avoid washing-out.
13038 The value must be a float number in the range [0,1] and defaults to
13039 0.210.
13040 antibanding
13042 Set the antibanding level. If enabled the filter will randomly vary
13043 the luminance of output pixels by a small amount to avoid banding
13044 of the histogram. Possible values are "none", "weak" or "strong".
13045 It defaults to "none".
13046 histogram
13048 Compute and draw a color distribution histogram for the input video.
13049 The computed histogram is a representation of the color component
13051 distribution in an image.
13052 Standard histogram displays the color components distribution in an
13054 image. Displays color graph for each color component. Shows
13055 distribution of the Y, U, V, A or R, G, B components, depending on
13056 input format, in the current frame. Below each graph a color component
13057 scale meter is shown.
13058 The filter accepts the following options:
13060 level_height
13062 Set height of level. Default value is 200. Allowed range is [50,
13063 2048].
13064 scale_height
13066 Set height of color scale. Default value is 12. Allowed range is
13067 [0, 40].
13068 display_mode
13070 Set display mode. It accepts the following values:
13071 stack
13073 Per color component graphs are placed below each other.
13074 parade
13076 Per color component graphs are placed side by side.
13077 overlay
13079 Presents information identical to that in the "parade", except
13080 that the graphs representing color components are superimposed
13081 directly over one another.
13082 Default is "stack".
13084 levels_mode
13086 Set mode. Can be either "linear", or "logarithmic". Default is
13087 "linear".
13088 components
13090 Set what color components to display. Default is 7.
13091 fgopacity
13093 Set foreground opacity. Default is 0.7.
13094 bgopacity
13096 Set background opacity. Default is 0.5.
13097 colors_mode
13099 Set colors mode. It accepts the following values:
13100 whiteonblack
13102 blackonwhite
13103 whiteongray
13104 blackongray
13105 coloronblack
13106 coloronwhite
13107 colorongray
13108 blackoncolor
13109 whiteoncolor
13110 grayoncolor
13111 Default is "whiteonblack".
13113 Examples
13115 • Calculate and draw histogram:
13117 ffplay -i input -vf histogram
13119 hqdn3d
13121 This is a high precision/quality 3d denoise filter. It aims to reduce
13122 image noise, producing smooth images and making still images really
13123 still. It should enhance compressibility.
13124 It accepts the following optional parameters:
13126 luma_spatial
13128 A non-negative floating point number which specifies spatial luma
13129 strength. It defaults to 4.0.
13130 chroma_spatial
13132 A non-negative floating point number which specifies spatial chroma
13133 strength. It defaults to 3.0*luma_spatial/4.0.
13134 luma_tmp
13136 A floating point number which specifies luma temporal strength. It
13137 defaults to 6.0*luma_spatial/4.0.
13138 chroma_tmp
13140 A floating point number which specifies chroma temporal strength.
13141 It defaults to luma_tmp*chroma_spatial/luma_spatial.
13142 Commands
13144 This filter supports same commands as options. The command accepts the
13146 same syntax of the corresponding option.
13147 If the specified expression is not valid, it is kept at its current
13149 value.
13150 hwdownload
13152 Download hardware frames to system memory.
13153 The input must be in hardware frames, and the output a non-hardware
13155 format. Not all formats will be supported on the output - it may be
13156 necessary to insert an additional format filter immediately following
13157 in the graph to get the output in a supported format.
13158 hwmap
13160 Map hardware frames to system memory or to another device.
13161 This filter has several different modes of operation; which one is used
13163 depends on the input and output formats:
13164 • Hardware frame input, normal frame output
13166 Map the input frames to system memory and pass them to the output.
13168 If the original hardware frame is later required (for example,
13169 after overlaying something else on part of it), the hwmap filter
13170 can be used again in the next mode to retrieve it.
13171 • Normal frame input, hardware frame output
13173 If the input is actually a software-mapped hardware frame, then
13175 unmap it - that is, return the original hardware frame.
13176 Otherwise, a device must be provided. Create new hardware surfaces
13178 on that device for the output, then map them back to the software
13179 format at the input and give those frames to the preceding filter.
13180 This will then act like the hwupload filter, but may be able to
13181 avoid an additional copy when the input is already in a compatible
13182 format.
13183 • Hardware frame input and output
13185 A device must be supplied for the output, either directly or with
13187 the derive_device option. The input and output devices must be of
13188 different types and compatible - the exact meaning of this is
13189 system-dependent, but typically it means that they must refer to
13190 the same underlying hardware context (for example, refer to the
13191 same graphics card).
13192 If the input frames were originally created on the output device,
13194 then unmap to retrieve the original frames.
13195 Otherwise, map the frames to the output device - create new
13197 hardware frames on the output corresponding to the frames on the
13198 input.
13199 The following additional parameters are accepted:
13201 mode
13203 Set the frame mapping mode. Some combination of:
13204 read
13206 The mapped frame should be readable.
13207 write
13209 The mapped frame should be writeable.
13210 overwrite
13212 The mapping will always overwrite the entire frame.
13213 This may improve performance in some cases, as the original
13215 contents of the frame need not be loaded.
13216 direct
13218 The mapping must not involve any copying.
13219 Indirect mappings to copies of frames are created in some cases
13221 where either direct mapping is not possible or it would have
13222 unexpected properties. Setting this flag ensures that the
13223 mapping is direct and will fail if that is not possible.
13224 Defaults to read+write if not specified.
13226 derive_device type
13228 Rather than using the device supplied at initialisation, instead
13229 derive a new device of type type from the device the input frames
13230 exist on.
13231 reverse
13233 In a hardware to hardware mapping, map in reverse - create frames
13234 in the sink and map them back to the source. This may be necessary
13235 in some cases where a mapping in one direction is required but only
13236 the opposite direction is supported by the devices being used.
13237 This option is dangerous - it may break the preceding filter in
13239 undefined ways if there are any additional constraints on that
13240 filter's output. Do not use it without fully understanding the
13241 implications of its use.
13242 hwupload
13244 Upload system memory frames to hardware surfaces.
13245 The device to upload to must be supplied when the filter is
13247 initialised. If using ffmpeg, select the appropriate device with the
13248 -filter_hw_device option or with the derive_device option. The input
13249 and output devices must be of different types and compatible - the
13250 exact meaning of this is system-dependent, but typically it means that
13251 they must refer to the same underlying hardware context (for example,
13252 refer to the same graphics card).
13253 The following additional parameters are accepted:
13255 derive_device type
13257 Rather than using the device supplied at initialisation, instead
13258 derive a new device of type type from the device the input frames
13259 exist on.
13260 hwupload_cuda
13262 Upload system memory frames to a CUDA device.
13263 It accepts the following optional parameters:
13265 device
13267 The number of the CUDA device to use
13268 hqx
13270 Apply a high-quality magnification filter designed for pixel art. This
13271 filter was originally created by Maxim Stepin.
13272 It accepts the following option:
13274 n Set the scaling dimension: 2 for "hq2x", 3 for "hq3x" and 4 for
13276 "hq4x". Default is 3.
13277 hstack
13279 Stack input videos horizontally.
13280 All streams must be of same pixel format and of same height.
13282 Note that this filter is faster than using overlay and pad filter to
13284 create same output.
13285 The filter accepts the following option:
13287 inputs
13289 Set number of input streams. Default is 2.
13290 shortest
13292 If set to 1, force the output to terminate when the shortest input
13293 terminates. Default value is 0.
13294 hsvhold
13296 Turns a certain HSV range into gray values.
13297 This filter measures color difference between set HSV color in options
13299 and ones measured in video stream. Depending on options, output colors
13300 can be changed to be gray or not.
13301 The filter accepts the following options:
13303 hue Set the hue value which will be used in color difference
13305 calculation. Allowed range is from -360 to 360. Default value is
13306 0.
13307 sat Set the saturation value which will be used in color difference
13309 calculation. Allowed range is from -1 to 1. Default value is 0.
13310 val Set the value which will be used in color difference calculation.
13312 Allowed range is from -1 to 1. Default value is 0.
13313 similarity
13315 Set similarity percentage with the key color. Allowed range is
13316 from 0 to 1. Default value is 0.01.
13317 0.00001 matches only the exact key color, while 1.0 matches
13319 everything.
13320 blend
13322 Blend percentage. Allowed range is from 0 to 1. Default value is
13323 0.
13324 0.0 makes pixels either fully gray, or not gray at all.
13326 Higher values result in more gray pixels, with a higher gray pixel
13328 the more similar the pixels color is to the key color.
13329 hsvkey
13331 Turns a certain HSV range into transparency.
13332 This filter measures color difference between set HSV color in options
13334 and ones measured in video stream. Depending on options, output colors
13335 can be changed to transparent by adding alpha channel.
13336 The filter accepts the following options:
13338 hue Set the hue value which will be used in color difference
13340 calculation. Allowed range is from -360 to 360. Default value is
13341 0.
13342 sat Set the saturation value which will be used in color difference
13344 calculation. Allowed range is from -1 to 1. Default value is 0.
13345 val Set the value which will be used in color difference calculation.
13347 Allowed range is from -1 to 1. Default value is 0.
13348 similarity
13350 Set similarity percentage with the key color. Allowed range is
13351 from 0 to 1. Default value is 0.01.
13352 0.00001 matches only the exact key color, while 1.0 matches
13354 everything.
13355 blend
13357 Blend percentage. Allowed range is from 0 to 1. Default value is
13358 0.
13359 0.0 makes pixels either fully transparent, or not transparent at
13361 all.
13362 Higher values result in semi-transparent pixels, with a higher
13364 transparency the more similar the pixels color is to the key color.
13365 hue
13367 Modify the hue and/or the saturation of the input.
13368 It accepts the following parameters:
13370 h Specify the hue angle as a number of degrees. It accepts an
13372 expression, and defaults to "0".
13373 s Specify the saturation in the [-10,10] range. It accepts an
13375 expression and defaults to "1".
13376 H Specify the hue angle as a number of radians. It accepts an
13378 expression, and defaults to "0".
13379 b Specify the brightness in the [-10,10] range. It accepts an
13381 expression and defaults to "0".
13382 h and H are mutually exclusive, and can't be specified at the same
13384 time.
13385 The b, h, H and s option values are expressions containing the
13387 following constants:
13388 n frame count of the input frame starting from 0
13390 pts presentation timestamp of the input frame expressed in time base
13392 units
13393 r frame rate of the input video, NAN if the input frame rate is
13395 unknown
13396 t timestamp expressed in seconds, NAN if the input timestamp is
13398 unknown
13399 tb time base of the input video
13401 Examples
13403 • Set the hue to 90 degrees and the saturation to 1.0:
13405 hue=h=90:s=1
13407 • Same command but expressing the hue in radians:
13409 hue=H=PI/2:s=1
13411 • Rotate hue and make the saturation swing between 0 and 2 over a
13413 period of 1 second:
13414 hue="H=2*PI*t: s=sin(2*PI*t)+1"
13416 • Apply a 3 seconds saturation fade-in effect starting at 0:
13418 hue="s=min(t/3\,1)"
13420 The general fade-in expression can be written as:
13422 hue="s=min(0\, max((t-START)/DURATION\, 1))"
13424 • Apply a 3 seconds saturation fade-out effect starting at 5 seconds:
13426 hue="s=max(0\, min(1\, (8-t)/3))"
13428 The general fade-out expression can be written as:
13430 hue="s=max(0\, min(1\, (START+DURATION-t)/DURATION))"
13432 Commands
13434 This filter supports the following commands:
13436 b
13438 s
13439 h
13440 H Modify the hue and/or the saturation and/or brightness of the input
13441 video. The command accepts the same syntax of the corresponding
13442 option.
13443 If the specified expression is not valid, it is kept at its current
13445 value.
13446 huesaturation
13448 Apply hue-saturation-intensity adjustments to input video stream.
13449 This filter operates in RGB colorspace.
13451 This filter accepts the following options:
13453 hue Set the hue shift in degrees to apply. Default is 0. Allowed range
13455 is from -180 to 180.
13456 saturation
13458 Set the saturation shift. Default is 0. Allowed range is from -1
13459 to 1.
13460 intensity
13462 Set the intensity shift. Default is 0. Allowed range is from -1 to
13463 1.
13464 colors
13466 Set which primary and complementary colors are going to be
13467 adjusted. This options is set by providing one or multiple values.
13468 This can select multiple colors at once. By default all colors are
13469 selected.
13470 r Adjust reds.
13472 y Adjust yellows.
13474 g Adjust greens.
13476 c Adjust cyans.
13478 b Adjust blues.
13480 m Adjust magentas.
13482 a Adjust all colors.
13484 strength
13486 Set strength of filtering. Allowed range is from 0 to 100. Default
13487 value is 1.
13488 rw, gw, bw
13490 Set weight for each RGB component. Allowed range is from 0 to 1.
13491 By default is set to 0.333, 0.334, 0.333. Those options are used
13492 in saturation and lightess processing.
13493 lightness
13495 Set preserving lightness, by default is disabled. Adjusting hues
13496 can change lightness from original RGB triplet, with this option
13497 enabled lightness is kept at same value.
13498 hysteresis
13500 Grow first stream into second stream by connecting components. This
13501 makes it possible to build more robust edge masks.
13502 This filter accepts the following options:
13504 planes
13506 Set which planes will be processed as bitmap, unprocessed planes
13507 will be copied from first stream. By default value 0xf, all planes
13508 will be processed.
13509 threshold
13511 Set threshold which is used in filtering. If pixel component value
13512 is higher than this value filter algorithm for connecting
13513 components is activated. By default value is 0.
13514 The "hysteresis" filter also supports the framesync options.
13516 iccdetect
13518 Detect the colorspace from an embedded ICC profile (if present), and
13519 update the frame's tags accordingly.
13520 This filter accepts the following options:
13522 force
13524 If true, the frame's existing colorspace tags will always be
13525 overridden by values detected from an ICC profile. Otherwise, they
13526 will only be assigned if they contain "unknown". Enabled by
13527 default.
13528 iccgen
13530 Generate ICC profiles and attach them to frames.
13531 This filter accepts the following options:
13533 color_primaries
13535 color_trc
13536 Configure the colorspace that the ICC profile will be generated
13537 for. The default value of "auto" infers the value from the input
13538 frame's metadata, defaulting to BT.709/sRGB as appropriate.
13539 See the setparams filter for a list of possible values, but note
13541 that "unknown" are not valid values for this filter.
13542 force
13544 If true, an ICC profile will be generated even if it would
13545 overwrite an already existing ICC profile. Disabled by default.
13546 identity
13548 Obtain the identity score between two input videos.
13549 This filter takes two input videos.
13551 Both input videos must have the same resolution and pixel format for
13553 this filter to work correctly. Also it assumes that both inputs have
13554 the same number of frames, which are compared one by one.
13555 The obtained per component, average, min and max identity score is
13557 printed through the logging system.
13558 The filter stores the calculated identity scores of each frame in frame
13560 metadata.
13561 This filter also supports the framesync options.
13563 In the below example the input file main.mpg being processed is
13565 compared with the reference file ref.mpg.
13566 ffmpeg -i main.mpg -i ref.mpg -lavfi identity -f null -
13568 idet
13570 Detect video interlacing type.
13571 This filter tries to detect if the input frames are interlaced,
13573 progressive, top or bottom field first. It will also try to detect
13574 fields that are repeated between adjacent frames (a sign of telecine).
13575 Single frame detection considers only immediately adjacent frames when
13577 classifying each frame. Multiple frame detection incorporates the
13578 classification history of previous frames.
13579 The filter will log these metadata values:
13581 single.current_frame
13583 Detected type of current frame using single-frame detection. One
13584 of: ``tff'' (top field first), ``bff'' (bottom field first),
13585 ``progressive'', or ``undetermined''
13586 single.tff
13588 Cumulative number of frames detected as top field first using
13589 single-frame detection.
13590 multiple.tff
13592 Cumulative number of frames detected as top field first using
13593 multiple-frame detection.
13594 single.bff
13596 Cumulative number of frames detected as bottom field first using
13597 single-frame detection.
13598 multiple.current_frame
13600 Detected type of current frame using multiple-frame detection. One
13601 of: ``tff'' (top field first), ``bff'' (bottom field first),
13602 ``progressive'', or ``undetermined''
13603 multiple.bff
13605 Cumulative number of frames detected as bottom field first using
13606 multiple-frame detection.
13607 single.progressive
13609 Cumulative number of frames detected as progressive using single-
13610 frame detection.
13611 multiple.progressive
13613 Cumulative number of frames detected as progressive using multiple-
13614 frame detection.
13615 single.undetermined
13617 Cumulative number of frames that could not be classified using
13618 single-frame detection.
13619 multiple.undetermined
13621 Cumulative number of frames that could not be classified using
13622 multiple-frame detection.
13623 repeated.current_frame
13625 Which field in the current frame is repeated from the last. One of
13626 ``neither'', ``top'', or ``bottom''.
13627 repeated.neither
13629 Cumulative number of frames with no repeated field.
13630 repeated.top
13632 Cumulative number of frames with the top field repeated from the
13633 previous frame's top field.
13634 repeated.bottom
13636 Cumulative number of frames with the bottom field repeated from the
13637 previous frame's bottom field.
13638 The filter accepts the following options:
13640 intl_thres
13642 Set interlacing threshold.
13643 prog_thres
13645 Set progressive threshold.
13646 rep_thres
13648 Threshold for repeated field detection.
13649 half_life
13651 Number of frames after which a given frame's contribution to the
13652 statistics is halved (i.e., it contributes only 0.5 to its
13653 classification). The default of 0 means that all frames seen are
13654 given full weight of 1.0 forever.
13655 analyze_interlaced_flag
13657 When this is not 0 then idet will use the specified number of
13658 frames to determine if the interlaced flag is accurate, it will not
13659 count undetermined frames. If the flag is found to be accurate it
13660 will be used without any further computations, if it is found to be
13661 inaccurate it will be cleared without any further computations.
13662 This allows inserting the idet filter as a low computational method
13663 to clean up the interlaced flag
13664 il
13666 Deinterleave or interleave fields.
13667 This filter allows one to process interlaced images fields without
13669 deinterlacing them. Deinterleaving splits the input frame into 2 fields
13670 (so called half pictures). Odd lines are moved to the top half of the
13671 output image, even lines to the bottom half. You can process (filter)
13672 them independently and then re-interleave them.
13673 The filter accepts the following options:
13675 luma_mode, l
13677 chroma_mode, c
13678 alpha_mode, a
13679 Available values for luma_mode, chroma_mode and alpha_mode are:
13680 none
13682 Do nothing.
13683 deinterleave, d
13685 Deinterleave fields, placing one above the other.
13686 interleave, i
13688 Interleave fields. Reverse the effect of deinterleaving.
13689 Default value is "none".
13691 luma_swap, ls
13693 chroma_swap, cs
13694 alpha_swap, as
13695 Swap luma/chroma/alpha fields. Exchange even & odd lines. Default
13696 value is 0.
13697 Commands
13699 This filter supports the all above options as commands.
13701 inflate
13703 Apply inflate effect to the video.
13704 This filter replaces the pixel by the local(3x3) average by taking into
13706 account only values higher than the pixel.
13707 It accepts the following options:
13709 threshold0
13711 threshold1
13712 threshold2
13713 threshold3
13714 Limit the maximum change for each plane, default is 65535. If 0,
13715 plane will remain unchanged.
13716 Commands
13718 This filter supports the all above options as commands.
13720 interlace
13722 Simple interlacing filter from progressive contents. This interleaves
13723 upper (or lower) lines from odd frames with lower (or upper) lines from
13724 even frames, halving the frame rate and preserving image height.
13725 Original Original New Frame
13727 Frame 'j' Frame 'j+1' (tff)
13728 ========== =========== ==================
13729 Line 0 --------------------> Frame 'j' Line 0
13730 Line 1 Line 1 ----> Frame 'j+1' Line 1
13731 Line 2 ---------------------> Frame 'j' Line 2
13732 Line 3 Line 3 ----> Frame 'j+1' Line 3
13733 ... ... ...
13734 New Frame + 1 will be generated by Frame 'j+2' and Frame 'j+3' and so on
13735 It accepts the following optional parameters:
13737 scan
13739 This determines whether the interlaced frame is taken from the even
13740 (tff - default) or odd (bff) lines of the progressive frame.
13741 lowpass
13743 Vertical lowpass filter to avoid twitter interlacing and reduce
13744 moire patterns.
13745 0, off
13747 Disable vertical lowpass filter
13748 1, linear
13750 Enable linear filter (default)
13751 2, complex
13753 Enable complex filter. This will slightly less reduce twitter
13754 and moire but better retain detail and subjective sharpness
13755 impression.
13756 kerndeint
13758 Deinterlace input video by applying Donald Graft's adaptive kernel
13759 deinterling. Work on interlaced parts of a video to produce progressive
13760 frames.
13761 The description of the accepted parameters follows.
13763 thresh
13765 Set the threshold which affects the filter's tolerance when
13766 determining if a pixel line must be processed. It must be an
13767 integer in the range [0,255] and defaults to 10. A value of 0 will
13768 result in applying the process on every pixels.
13769 map Paint pixels exceeding the threshold value to white if set to 1.
13771 Default is 0.
13772 order
13774 Set the fields order. Swap fields if set to 1, leave fields alone
13775 if 0. Default is 0.
13776 sharp
13778 Enable additional sharpening if set to 1. Default is 0.
13779 twoway
13781 Enable twoway sharpening if set to 1. Default is 0.
13782 Examples
13784 • Apply default values:
13786 kerndeint=thresh=10:map=0:order=0:sharp=0:twoway=0
13788 • Enable additional sharpening:
13790 kerndeint=sharp=1
13792 • Paint processed pixels in white:
13794 kerndeint=map=1
13796 kirsch
13798 Apply kirsch operator to input video stream.
13799 The filter accepts the following option:
13801 planes
13803 Set which planes will be processed, unprocessed planes will be
13804 copied. By default value 0xf, all planes will be processed.
13805 scale
13807 Set value which will be multiplied with filtered result.
13808 delta
13810 Set value which will be added to filtered result.
13811 Commands
13813 This filter supports the all above options as commands.
13815 lagfun
13817 Slowly update darker pixels.
13818 This filter makes short flashes of light appear longer. This filter
13820 accepts the following options:
13821 decay
13823 Set factor for decaying. Default is .95. Allowed range is from 0 to
13824 1.
13825 planes
13827 Set which planes to filter. Default is all. Allowed range is from 0
13828 to 15.
13829 Commands
13831 This filter supports the all above options as commands.
13833 lenscorrection
13835 Correct radial lens distortion
13836 This filter can be used to correct for radial distortion as can result
13838 from the use of wide angle lenses, and thereby re-rectify the image. To
13839 find the right parameters one can use tools available for example as
13840 part of opencv or simply trial-and-error. To use opencv use the
13841 calibration sample (under samples/cpp) from the opencv sources and
13842 extract the k1 and k2 coefficients from the resulting matrix.
13843 Note that effectively the same filter is available in the open-source
13845 tools Krita and Digikam from the KDE project.
13846 In contrast to the vignette filter, which can also be used to
13848 compensate lens errors, this filter corrects the distortion of the
13849 image, whereas vignette corrects the brightness distribution, so you
13850 may want to use both filters together in certain cases, though you will
13851 have to take care of ordering, i.e. whether vignetting should be
13852 applied before or after lens correction.
13853 Options
13855 The filter accepts the following options:
13857 cx Relative x-coordinate of the focal point of the image, and thereby
13859 the center of the distortion. This value has a range [0,1] and is
13860 expressed as fractions of the image width. Default is 0.5.
13861 cy Relative y-coordinate of the focal point of the image, and thereby
13863 the center of the distortion. This value has a range [0,1] and is
13864 expressed as fractions of the image height. Default is 0.5.
13865 k1 Coefficient of the quadratic correction term. This value has a
13867 range [-1,1]. 0 means no correction. Default is 0.
13868 k2 Coefficient of the double quadratic correction term. This value has
13870 a range [-1,1]. 0 means no correction. Default is 0.
13871 i Set interpolation type. Can be "nearest" or "bilinear". Default is
13873 "nearest".
13874 fc Specify the color of the unmapped pixels. For the syntax of this
13876 option, check the "Color" section in the ffmpeg-utils manual.
13877 Default color is "black@0".
13878 The formula that generates the correction is:
13880 r_src = r_tgt * (1 + k1 * (r_tgt / r_0)^2 + k2 * (r_tgt / r_0)^4)
13882 where r_0 is halve of the image diagonal and r_src and r_tgt are the
13884 distances from the focal point in the source and target images,
13885 respectively.
13886 Commands
13888 This filter supports the all above options as commands.
13890 lensfun
13892 Apply lens correction via the lensfun library
13893 (<http://lensfun.sourceforge.net/>).
13894 The "lensfun" filter requires the camera make, camera model, and lens
13896 model to apply the lens correction. The filter will load the lensfun
13897 database and query it to find the corresponding camera and lens entries
13898 in the database. As long as these entries can be found with the given
13899 options, the filter can perform corrections on frames. Note that
13900 incomplete strings will result in the filter choosing the best match
13901 with the given options, and the filter will output the chosen camera
13902 and lens models (logged with level "info"). You must provide the make,
13903 camera model, and lens model as they are required.
13904 To obtain a list of available makes and models, leave out one or both
13906 of "make" and "model" options. The filter will send the full list to
13907 the log with level "INFO". The first column is the make and the second
13908 column is the model. To obtain a list of available lenses, set any
13909 values for make and model and leave out the "lens_model" option. The
13910 filter will send the full list of lenses in the log with level "INFO".
13911 The ffmpeg tool will exit after the list is printed.
13912 The filter accepts the following options:
13914 make
13916 The make of the camera (for example, "Canon"). This option is
13917 required.
13918 model
13920 The model of the camera (for example, "Canon EOS 100D"). This
13921 option is required.
13922 lens_model
13924 The model of the lens (for example, "Canon EF-S 18-55mm f/3.5-5.6
13925 IS STM"). This option is required.
13926 db_path
13928 The full path to the lens database folder. If not set, the filter
13929 will attempt to load the database from the install path when the
13930 library was built. Default is unset.
13931 mode
13933 The type of correction to apply. The following values are valid
13934 options:
13935 vignetting
13937 Enables fixing lens vignetting.
13938 geometry
13940 Enables fixing lens geometry. This is the default.
13941 subpixel
13943 Enables fixing chromatic aberrations.
13944 vig_geo
13946 Enables fixing lens vignetting and lens geometry.
13947 vig_subpixel
13949 Enables fixing lens vignetting and chromatic aberrations.
13950 distortion
13952 Enables fixing both lens geometry and chromatic aberrations.
13953 all Enables all possible corrections.
13955 focal_length
13957 The focal length of the image/video (zoom; expected constant for
13958 video). For example, a 18--55mm lens has focal length range of
13959 [18--55], so a value in that range should be chosen when using that
13960 lens. Default 18.
13961 aperture
13963 The aperture of the image/video (expected constant for video). Note
13964 that aperture is only used for vignetting correction. Default 3.5.
13965 focus_distance
13967 The focus distance of the image/video (expected constant for
13968 video). Note that focus distance is only used for vignetting and
13969 only slightly affects the vignetting correction process. If
13970 unknown, leave it at the default value (which is 1000).
13971 scale
13973 The scale factor which is applied after transformation. After
13974 correction the video is no longer necessarily rectangular. This
13975 parameter controls how much of the resulting image is visible. The
13976 value 0 means that a value will be chosen automatically such that
13977 there is little or no unmapped area in the output image. 1.0 means
13978 that no additional scaling is done. Lower values may result in more
13979 of the corrected image being visible, while higher values may avoid
13980 unmapped areas in the output.
13981 target_geometry
13983 The target geometry of the output image/video. The following values
13984 are valid options:
13985 rectilinear (default)
13987 fisheye
13988 panoramic
13989 equirectangular
13990 fisheye_orthographic
13991 fisheye_stereographic
13992 fisheye_equisolid
13993 fisheye_thoby
13994 reverse
13995 Apply the reverse of image correction (instead of correcting
13996 distortion, apply it).
13997 interpolation
13999 The type of interpolation used when correcting distortion. The
14000 following values are valid options:
14001 nearest
14003 linear (default)
14004 lanczos
14005 Examples
14007 • Apply lens correction with make "Canon", camera model "Canon EOS
14009 100D", and lens model "Canon EF-S 18-55mm f/3.5-5.6 IS STM" with
14010 focal length of "18" and aperture of "8.0".
14011 ffmpeg -i input.mov -vf lensfun=make=Canon:model="Canon EOS 100D":lens_model="Canon EF-S 18-55mm f/3.5-5.6 IS STM":focal_length=18:aperture=8 -c:v h264 -b:v 8000k output.mov
14013 • Apply the same as before, but only for the first 5 seconds of
14015 video.
14016 ffmpeg -i input.mov -vf lensfun=make=Canon:model="Canon EOS 100D":lens_model="Canon EF-S 18-55mm f/3.5-5.6 IS STM":focal_length=18:aperture=8:enable='lte(t\,5)' -c:v h264 -b:v 8000k output.mov
14018 libplacebo
14020 Flexible GPU-accelerated processing filter based on libplacebo
14021 (<https://code.videolan.org/videolan/libplacebo>). Note that this
14022 filter currently only accepts Vulkan input frames.
14023 Options
14025 The options for this filter are divided into the following sections:
14027 Output mode
14029 These options control the overall output mode. By default, libplacebo
14031 will try to preserve the source colorimetry and size as best as it can,
14032 but it will apply any embedded film grain, dolby vision metadata or
14033 anamorphic SAR present in source frames.
14034 w
14036 h Set the output video dimension expression. Default value is the
14037 input dimension.
14038 Allows for the same expressions as the scale filter.
14040 format
14042 Set the output format override. If unset (the default), frames will
14043 be output in the same format as the respective input frames.
14044 Otherwise, format conversion will be performed.
14045 force_original_aspect_ratio
14047 force_divisible_by
14048 Work the same as the identical scale filter options.
14049 normalize_sar
14051 If enabled, output frames will always have a pixel aspect ratio of
14052 1:1. This will introduce padding/cropping as necessary. If disabled
14053 (the default), any aspect ratio mismatches, including those from
14054 e.g. anamorphic video sources, are forwarded to the output pixel
14055 aspect ratio.
14056 pad_crop_ratio
14058 Specifies a ratio (between 0.0 and 1.0) between padding and
14059 cropping when the input aspect ratio does not match the output
14060 aspect ratio and normalize_sar is in effect. The default of 0.0
14061 always pads the content with black borders, while a value of 1.0
14062 always crops off parts of the content. Intermediate values are
14063 possible, leading to a mix of the two approaches.
14064 colorspace
14066 color_primaries
14067 color_trc
14068 range
14069 Configure the colorspace that output frames will be delivered in.
14070 The default value of "auto" outputs frames in the same format as
14071 the input frames, leading to no change. For any other value,
14072 conversion will be performed.
14073 See the setparams filter for a list of possible values.
14075 apply_filmgrain
14077 Apply film grain (e.g. AV1 or H.274) if present in source frames,
14078 and strip it from the output. Enabled by default.
14079 apply_dolbyvision
14081 Apply Dolby Vision RPU metadata if present in source frames, and
14082 strip it from the output. Enabled by default. Note that Dolby
14083 Vision will always output BT.2020+PQ, overriding the usual input
14084 frame metadata. These will also be picked as the values of "auto"
14085 for the respective frame output options.
14086 Scaling
14088 The options in this section control how libplacebo performs upscaling
14090 and (if necessary) downscaling. Note that libplacebo will always
14091 internally operate on 4:4:4 content, so any sub-sampled chroma formats
14092 such as "yuv420p" will necessarily be upsampled and downsampled as part
14093 of the rendering process. That means scaling might be in effect even if
14094 the source and destination resolution are the same.
14095 upscaler
14097 downscaler
14098 Configure the filter kernel used for upscaling and downscaling. The
14099 respective defaults are "spline36" and "mitchell". For a full list
14100 of possible values, pass "help" to these options. The most
14101 important values are:
14102 none
14104 Forces the use of built-in GPU texture sampling (typically
14105 bilinear). Extremely fast but poor quality, especially when
14106 downscaling.
14107 bilinear
14109 Bilinear interpolation. Can generally be done for free on GPUs,
14110 except when doing so would lead to aliasing. Fast and low
14111 quality.
14112 nearest
14114 Nearest-neighbour interpolation. Sharp but highly aliasing.
14115 oversample
14117 Algorithm that looks visually similar to nearest-neighbour
14118 interpolation but tries to preserve pixel aspect ratio. Good
14119 for pixel art, since it results in minimal distortion of the
14120 artistic appearance.
14121 lanczos
14123 Standard sinc-sinc interpolation kernel.
14124 spline36
14126 Cubic spline approximation of lanczos. No difference in
14127 performance, but has very slightly less ringing.
14128 ewa_lanczos
14130 Elliptically weighted average version of lanczos, based on a
14131 jinc-sinc kernel. This is also popularly referred to as just
14132 "Jinc scaling". Slow but very high quality.
14133 gaussian
14135 Gaussian kernel. Has certain ideal mathematical properties, but
14136 subjectively very blurry.
14137 mitchell
14139 Cubic BC spline with parameters recommended by Mitchell and
14140 Netravali. Very little ringing.
14141 lut_entries
14143 Configures the size of scaler LUTs, ranging from 1 to 256. The
14144 default of 0 will pick libplacebo's internal default, typically 64.
14145 antiringing
14147 Enables anti-ringing (for non-EWA filters). The value (between 0.0
14148 and 1.0) configures the strength of the anti-ringing algorithm. May
14149 increase aliasing if set too high. Disabled by default.
14150 sigmoid
14152 Enable sigmoidal compression during upscaling. Reduces ringing
14153 slightly. Enabled by default.
14154 Debanding
14156 Libplacebo comes with a built-in debanding filter that is good at
14158 counteracting many common sources of banding and blocking. Turning this
14159 on is highly recommended whenever quality is desired.
14160 deband
14162 Enable (fast) debanding algorithm. Disabled by default.
14163 deband_iterations
14165 Number of deband iterations of the debanding algorithm. Each
14166 iteration is performed with progressively increased radius (and
14167 diminished threshold). Recommended values are in the range 1 to 4.
14168 Defaults to 1.
14169 deband_threshold
14171 Debanding filter strength. Higher numbers lead to more aggressive
14172 debanding. Defaults to 4.0.
14173 deband_radius
14175 Debanding filter radius. A higher radius is better for slow
14176 gradients, while a lower radius is better for steep gradients.
14177 Defaults to 16.0.
14178 deband_grain
14180 Amount of extra output grain to add. Helps hide imperfections.
14181 Defaults to 6.0.
14182 Color adjustment
14184 A collection of subjective color controls. Not very rigorous, so the
14186 exact effect will vary somewhat depending on the input primaries and
14187 colorspace.
14188 brightness
14190 Brightness boost, between -1.0 and 1.0. Defaults to 0.0.
14191 contrast
14193 Contrast gain, between 0.0 and 16.0. Defaults to 1.0.
14194 saturation
14196 Saturation gain, between 0.0 and 16.0. Defaults to 1.0.
14197 hue Hue shift in radians, between -3.14 and 3.14. Defaults to 0.0. This
14199 will rotate the UV subvector, defaulting to BT.709 coefficients for
14200 RGB inputs.
14201 gamma
14203 Gamma adjustment, between 0.0 and 16.0. Defaults to 1.0.
14204 cones
14206 Cone model to use for color blindness simulation. Accepts any
14207 combination of "l", "m" and "s". Here are some examples:
14208 m Deuteranomaly / deuteranopia (affecting 3%-4% of the
14210 population)
14211 l Protanomaly / protanopia (affecting 1%-2% of the population)
14213 l+m Monochromacy (very rare)
14215 l+m+s
14217 Achromatopsy (complete loss of daytime vision, extremely rare)
14218 cone-strength
14220 Gain factor for the cones specified by "cones", between 0.0 and
14221 10.0. A value of 1.0 results in no change to color vision. A value
14222 of 0.0 (the default) simulates complete loss of those cones. Values
14223 above 1.0 result in exaggerating the differences between cones,
14224 which may help compensate for reduced color vision.
14225 Peak detection
14227 To help deal with sources that only have static HDR10 metadata (or no
14229 tagging whatsoever), libplacebo uses its own internal frame analysis
14230 compute shader to analyze source frames and adapt the tone mapping
14231 function in realtime. If this is too slow, or if exactly reproducible
14232 frame-perfect results are needed, it's recommended to turn this feature
14233 off.
14234 peak_detect
14236 Enable HDR peak detection. Ignores static MaxCLL/MaxFALL values in
14237 favor of dynamic detection from the input. Note that the detected
14238 values do not get written back to the output frames, they merely
14239 guide the internal tone mapping process. Enabled by default.
14240 smoothing_period
14242 Peak detection smoothing period, between 0.0 and 1000.0. Higher
14243 values result in peak detection becoming less responsive to changes
14244 in the input. Defaults to 100.0.
14245 minimum_peak
14247 Lower bound on the detected peak (relative to SDR white), between
14248 0.0 and 100.0. Defaults to 1.0.
14249 scene_threshold_low
14251 scene_threshold_high
14252 Lower and upper thresholds for scene change detection. Expressed in
14253 a logarithmic scale between 0.0 and 100.0. Default to 5.5 and 10.0,
14254 respectively. Setting either to a negative value disables this
14255 functionality.
14256 overshoot
14258 Peak smoothing overshoot margin, between 0.0 and 1.0. Provides a
14259 safety margin to prevent clipping as a result of peak smoothing.
14260 Defaults to 0.05, corresponding to a margin of 5%.
14261 Tone mapping
14263 The options in this section control how libplacebo performs tone-
14265 mapping and gamut-mapping when dealing with mismatches between wide-
14266 gamut or HDR content. In general, libplacebo relies on accurate source
14267 tagging and mastering display gamut information to produce the best
14268 results.
14269 intent
14271 Rendering intent to use when adapting between different primary
14272 color gamuts (after tone-mapping).
14273 perceptual
14275 Perceptual gamut mapping. Currently equivalent to relative
14276 colorimetric.
14277 relative
14279 Relative colorimetric. This is the default.
14280 absolute
14282 Absolute colorimetric.
14283 saturation
14285 Saturation mapping. Forcibly stretches the source gamut to the
14286 target gamut.
14287 gamut_mode
14289 How to handle out-of-gamut colors that can occur as a result of
14290 colorimetric gamut mapping.
14291 clip
14293 Do nothing, simply clip out-of-range colors to the RGB volume.
14294 This is the default.
14295 warn
14297 Highlight out-of-gamut pixels (by coloring them pink).
14298 darken
14300 Linearly reduces content brightness to preserves saturated
14301 details, followed by clipping the remaining out-of-gamut
14302 colors. As the name implies, this makes everything darker, but
14303 provides a good balance between preserving details and colors.
14304 desaturate
14306 Hard-desaturates out-of-gamut colors towards white, while
14307 preserving the luminance. Has a tendency to shift colors.
14308 tonemapping
14310 Tone-mapping algorithm to use. Available values are:
14311 auto
14313 Automatic selection based on internal heuristics. This is the
14314 default.
14315 clip
14317 Performs no tone-mapping, just clips out-of-range colors.
14318 Retains perfect color accuracy for in-range colors but
14319 completely destroys out-of-range information. Does not perform
14320 any black point adaptation. Not configurable.
14321 st2094-40
14323 EETF from SMPTE ST 2094-40 Annex B, which applies the Bezier
14324 curves from HDR10+ dynamic metadata based on Bezier curves to
14325 perform tone-mapping. The OOTF used is adjusted based on the
14326 ratio between the targeted and actual display peak luminances.
14327 st2094-10
14329 EETF from SMPTE ST 2094-10 Annex B.2, which takes into account
14330 the input signal average luminance in addition to the
14331 maximum/minimum. The configurable contrast parameter influences
14332 the slope of the linear output segment, defaulting to 1.0 for
14333 no increase/decrease in contrast. Note that this does not
14334 currently include the subjective gain/offset/gamma controls
14335 defined in Annex B.3.
14336 bt.2390
14338 EETF from the ITU-R Report BT.2390, a hermite spline roll-off
14339 with linear segment. The knee point offset is configurable.
14340 Note that this parameter defaults to 1.0, rather than the value
14341 of 0.5 from the ITU-R spec.
14342 bt.2446a
14344 EETF from ITU-R Report BT.2446, method A. Designed for well-
14345 mastered HDR sources. Can be used for both forward and inverse
14346 tone mapping. Not configurable.
14347 spline
14349 Simple spline consisting of two polynomials, joined by a single
14350 pivot point. The parameter gives the pivot point (in PQ
14351 space), defaulting to 0.30. Can be used for both forward and
14352 inverse tone mapping.
14353 reinhard
14355 Simple non-linear, global tone mapping algorithm. The parameter
14356 specifies the local contrast coefficient at the display peak.
14357 Essentially, a parameter of 0.5 implies that the reference
14358 white will be about half as bright as when clipping. Defaults
14359 to 0.5, which results in the simplest formulation of this
14360 function.
14361 mobius
14363 Generalization of the reinhard tone mapping algorithm to
14364 support an additional linear slope near black. The tone mapping
14365 parameter indicates the trade-off between the linear section
14366 and the non-linear section. Essentially, for a given parameter
14367 x, every color value below x will be mapped linearly, while
14368 higher values get non-linearly tone-mapped. Values near 1.0
14369 make this curve behave like "clip", while values near 0.0 make
14370 this curve behave like "reinhard". The default value is 0.3,
14371 which provides a good balance between colorimetric accuracy and
14372 preserving out-of-gamut details.
14373 hable
14375 Piece-wise, filmic tone-mapping algorithm developed by John
14376 Hable for use in Uncharted 2, inspired by a similar tone-
14377 mapping algorithm used by Kodak. Popularized by its use in
14378 video games with HDR rendering. Preserves both dark and bright
14379 details very well, but comes with the drawback of changing the
14380 average brightness quite significantly. This is sort of similar
14381 to "reinhard" with parameter 0.24.
14382 gamma
14384 Fits a gamma (power) function to transfer between the source
14385 and target color spaces, effectively resulting in a perceptual
14386 hard-knee joining two roughly linear sections. This preserves
14387 details at all scales fairly accurately, but can result in an
14388 image with a muted or dull appearance. The parameter is used as
14389 the cutoff point, defaulting to 0.5.
14390 linear
14392 Linearly stretches the input range to the output range, in PQ
14393 space. This will preserve all details accurately, but results
14394 in a significantly different average brightness. Can be used
14395 for inverse tone-mapping in addition to regular tone-mapping.
14396 The parameter can be used as an additional linear gain
14397 coefficient (defaulting to 1.0).
14398 tonemapping_param
14400 For tunable tone mapping functions, this parameter can be used to
14401 fine-tune the curve behavior. Refer to the documentation of
14402 "tonemapping". The default value of 0.0 is replaced by the curve's
14403 preferred default setting.
14404 tonemapping_mode
14406 This option determines how the tone mapping function specified by
14407 "tonemapping" is applied to the colors in a scene. Possible values
14408 are:
14409 auto
14411 Automatic selection based on internal heuristics. This is the
14412 default.
14413 rgb Apply the function per-channel in the RGB colorspace. Per-
14415 channel tone-mapping in RGB. Guarantees no clipping and heavily
14416 desaturates the output, but distorts the colors quite
14417 significantly. Very similar to the "Hollywood" look and feel.
14418 max Tone-mapping is performed on the brightest component found in
14420 the signal. Good at preserving details in highlights, but has a
14421 tendency to crush blacks.
14422 hybrid
14424 Tone-map per-channel for highlights and linearly (luma-based)
14425 for midtones/shadows, based on a fixed gamma 2.4 coefficient
14426 curve.
14427 luma
14429 Tone-map linearly on the luma component (CIE Y), and adjust
14430 (desaturate) the chromaticities to compensate using a simple
14431 constant factor. This is essentially the mode used in ITU-R
14432 BT.2446 method A.
14433 inverse_tonemapping
14435 If enabled, this filter will also attempt stretching SDR signals to
14436 fill HDR output color volumes. Disabled by default.
14437 tonemapping_crosstalk
14439 Extra tone-mapping crosstalk factor, between 0.0 and 0.3. This can
14440 help reduce issues tone-mapping certain bright spectral colors.
14441 Defaults to 0.04.
14442 tonemapping_lut_size
14444 Size of the tone-mapping LUT, between 2 and 1024. Defaults to 256.
14445 Note that this figure is squared when combined with "peak_detect".
14446 Dithering
14448 By default, libplacebo will dither whenever necessary, which includes
14450 rendering to any integer format below 16-bit precision. It's
14451 recommended to always leave this on, since not doing so may result in
14452 visible banding in the output, even if the "debanding" filter is
14453 enabled. If maximum performance is needed, use "ordered_fixed" instead
14454 of disabling dithering.
14455 dithering
14457 Dithering method to use. Accepts the following values:
14458 none
14460 Disables dithering completely. May result in visible banding.
14461 blue
14463 Dither with pseudo-blue noise. This is the default.
14464 ordered
14466 Tunable ordered dither pattern.
14467 ordered_fixed
14469 Faster ordered dither with a fixed size of 6. Texture-less.
14470 white
14472 Dither with white noise. Texture-less.
14473 dither_lut_size
14475 Dither LUT size, as log base2 between 1 and 8. Defaults to 6,
14476 corresponding to a LUT size of "64x64".
14477 dither_temporal
14479 Enables temporal dithering. Disabled by default.
14480 Custom shaders
14482 libplacebo supports a number of custom shaders based on the mpv .hook
14484 GLSL syntax. A collection of such shaders can be found here:
14485 <https://github.com/mpv-player/mpv/wiki/User-Scripts#user-shaders>
14486 A full description of the mpv shader format is beyond the scope of this
14488 section, but a summary can be found here:
14489 <https://mpv.io/manual/master/#options-glsl-shader>
14490 custom_shader_path
14492 Specifies a path to a custom shader file to load at runtime.
14493 custom_shader_bin
14495 Specifies a complete custom shader as a raw string.
14496 Debugging / performance
14498 All of the options in this section default off. They may be of
14500 assistance when attempting to squeeze the maximum performance at the
14501 cost of quality.
14502 skip_aa
14504 Disable anti-aliasing when downscaling.
14505 polar_cutoff
14507 Truncate polar (EWA) scaler kernels below this absolute magnitude,
14508 between 0.0 and 1.0.
14509 disable_linear
14511 Disable linear light scaling.
14512 disable_builtin
14514 Disable built-in GPU sampling (forces LUT).
14515 disable_fbos
14517 Forcibly disable FBOs, resulting in loss of almost all
14518 functionality, but offering the maximum possible speed.
14519 Commands
14521 This filter supports almost all of the above options as commands.
14523 Examples
14525 • Complete example for how to initialize the Vulkan device, upload
14527 frames to the GPU, perform filter conversion to yuv420p, and
14528 download frames back to the CPU for output. Note that in specific
14529 cases you can get around the need to perform format conversion by
14530 specifying the correct "format" filter option corresponding to the
14531 input frames.
14532 ffmpeg -i $INPUT -init_hw_device vulkan -vf hwupload,libplacebo=format=yuv420p,hwdownload,format=yuv420p $OUTPUT
14534 • Tone-map input to standard gamut BT.709 output:
14536 libplacebo=colorspace=bt709:color_primaries=bt709:color_trc=bt709:range=tv
14538 • Rescale input to fit into standard 1080p, with high quality
14540 scaling:
14541 libplacebo=w=1920:h=1080:force_original_aspect_ratio=decrease:normalize_sar=true:upscaler=ewa_lanczos:downscaler=ewa_lanczos
14543 • Convert input to standard sRGB JPEG:
14545 libplacebo=format=yuv420p:colorspace=bt470bg:color_primaries=bt709:color_trc=iec61966-2-1:range=pc
14547 • Use higher quality debanding settings:
14549 libplacebo=deband=true:deband_iterations=3:deband_radius=8:deband_threshold=6
14551 • Run this filter on the CPU, on systems with Mesa installed (and
14553 with the most expensive options disabled):
14554 ffmpeg ... -init_hw_device vulkan:llvmpipe ... -vf libplacebo=upscaler=none:downscaler=none:peak_detect=false
14556 • Suppress CPU-based AV1/H.274 film grain application in the decoder,
14558 in favor of doing it with this filter. Note that this is only a
14559 gain if the frames are either already on the GPU, or if you're
14560 using libplacebo for other purposes, since otherwise the VRAM
14561 roundtrip will more than offset any expected speedup.
14562 ffmpeg -export_side_data +film_grain ... -vf libplacebo=apply_filmgrain=true
14564 libvmaf
14566 Calulate the VMAF (Video Multi-Method Assessment Fusion) score for a
14567 reference/distorted pair of input videos.
14568 The first input is the distorted video, and the second input is the
14570 reference video.
14571 The obtained VMAF score is printed through the logging system.
14573 It requires Netflix's vmaf library (libvmaf) as a pre-requisite. After
14575 installing the library it can be enabled using: "./configure
14576 --enable-libvmaf".
14577 The filter has following options:
14579 model
14581 A `|` delimited list of vmaf models. Each model can be configured
14582 with a number of parameters. Default value: "version=vmaf_v0.6.1"
14583 model_path
14585 Deprecated, use model='path=...'.
14586 enable_transform
14588 Deprecated, use model='enable_transform=true'.
14589 phone_model
14591 Deprecated, use model='enable_transform=true'.
14592 enable_conf_interval
14594 Deprecated, use model='enable_conf_interval=true'.
14595 feature
14597 A `|` delimited list of features. Each feature can be configured
14598 with a number of parameters.
14599 psnr
14601 Deprecated, use feature='name=psnr'.
14602 ssim
14604 Deprecated, use feature='name=ssim'.
14605 ms_ssim
14607 Deprecated, use feature='name=ms_ssim'.
14608 log_path
14610 Set the file path to be used to store log files.
14611 log_fmt
14613 Set the format of the log file (xml, json, csv, or sub).
14614 n_threads
14616 Set number of threads to be used when initializing libvmaf.
14617 Default value: 0, no threads.
14618 n_subsample
14620 Set frame subsampling interval to be used.
14621 This filter also supports the framesync options.
14623 Examples
14625 • In the examples below, a distorted video distorted.mpg is compared
14627 with a reference file reference.mpg.
14628 • Basic usage:
14630 ffmpeg -i distorted.mpg -i reference.mpg -lavfi libvmaf=log_path=output.xml -f null -
14632 • Example with multiple models:
14634 ffmpeg -i distorted.mpg -i reference.mpg -lavfi libvmaf='model=version=vmaf_v0.6.1\\:name=vmaf|version=vmaf_v0.6.1neg\\:name=vmaf_neg' -f null -
14636 • Example with multiple addtional features:
14638 ffmpeg -i distorted.mpg -i reference.mpg -lavfi libvmaf='feature=name=psnr|name=ciede' -f null -
14640 • Example with options and different containers:
14642 ffmpeg -i distorted.mpg -i reference.mkv -lavfi "[0:v]settb=AVTB,setpts=PTS-STARTPTS[main];[1:v]settb=AVTB,setpts=PTS-STARTPTS[ref];[main][ref]libvmaf=log_fmt=json:log_path=output.json" -f null -
14644 limitdiff
14646 Apply limited difference filter using second and optionally third video
14647 stream.
14648 The filter accepts the following options:
14650 threshold
14652 Set the threshold to use when allowing certain differences between
14653 video streams. Any absolute difference value lower or exact than
14654 this threshold will pick pixel components from first video stream.
14655 elasticity
14657 Set the elasticity of soft thresholding when processing video
14658 streams. This value multiplied with first one sets second
14659 threshold. Any absolute difference value greater or exact than
14660 second threshold will pick pixel components from second video
14661 stream. For values between those two threshold linear interpolation
14662 between first and second video stream will be used.
14663 reference
14665 Enable the reference (third) video stream processing. By default is
14666 disabled. If set, this video stream will be used for calculating
14667 absolute difference with first video stream.
14668 planes
14670 Specify which planes will be processed. Defaults to all available.
14671 Commands
14673 This filter supports the all above options as commands except option
14675 reference.
14676 limiter
14678 Limits the pixel components values to the specified range [min, max].
14679 The filter accepts the following options:
14681 min Lower bound. Defaults to the lowest allowed value for the input.
14683 max Upper bound. Defaults to the highest allowed value for the input.
14685 planes
14687 Specify which planes will be processed. Defaults to all available.
14688 Commands
14690 This filter supports the all above options as commands.
14692 loop
14694 Loop video frames.
14695 The filter accepts the following options:
14697 loop
14699 Set the number of loops. Setting this value to -1 will result in
14700 infinite loops. Default is 0.
14701 size
14703 Set maximal size in number of frames. Default is 0.
14704 start
14706 Set first frame of loop. Default is 0.
14707 Examples
14709 • Loop single first frame infinitely:
14711 loop=loop=-1:size=1:start=0
14713 • Loop single first frame 10 times:
14715 loop=loop=10:size=1:start=0
14717 • Loop 10 first frames 5 times:
14719 loop=loop=5:size=10:start=0
14721 lut1d
14723 Apply a 1D LUT to an input video.
14724 The filter accepts the following options:
14726 file
14728 Set the 1D LUT file name.
14729 Currently supported formats:
14731 cube
14733 Iridas
14734 csp cineSpace
14736 interp
14738 Select interpolation mode.
14739 Available values are:
14741 nearest
14743 Use values from the nearest defined point.
14744 linear
14746 Interpolate values using the linear interpolation.
14747 cosine
14749 Interpolate values using the cosine interpolation.
14750 cubic
14752 Interpolate values using the cubic interpolation.
14753 spline
14755 Interpolate values using the spline interpolation.
14756 Commands
14758 This filter supports the all above options as commands.
14760 lut3d
14762 Apply a 3D LUT to an input video.
14763 The filter accepts the following options:
14765 file
14767 Set the 3D LUT file name.
14768 Currently supported formats:
14770 3dl AfterEffects
14772 cube
14774 Iridas
14775 dat DaVinci
14777 m3d Pandora
14779 csp cineSpace
14781 interp
14783 Select interpolation mode.
14784 Available values are:
14786 nearest
14788 Use values from the nearest defined point.
14789 trilinear
14791 Interpolate values using the 8 points defining a cube.
14792 tetrahedral
14794 Interpolate values using a tetrahedron.
14795 pyramid
14797 Interpolate values using a pyramid.
14798 prism
14800 Interpolate values using a prism.
14801 Commands
14803 This filter supports the "interp" option as commands.
14805 lumakey
14807 Turn certain luma values into transparency.
14808 The filter accepts the following options:
14810 threshold
14812 Set the luma which will be used as base for transparency. Default
14813 value is 0.
14814 tolerance
14816 Set the range of luma values to be keyed out. Default value is
14817 0.01.
14818 softness
14820 Set the range of softness. Default value is 0. Use this to control
14821 gradual transition from zero to full transparency.
14822 Commands
14824 This filter supports same commands as options. The command accepts the
14826 same syntax of the corresponding option.
14827 If the specified expression is not valid, it is kept at its current
14829 value.
14830 lut, lutrgb, lutyuv
14832 Compute a look-up table for binding each pixel component input value to
14833 an output value, and apply it to the input video.
14834 lutyuv applies a lookup table to a YUV input video, lutrgb to an RGB
14836 input video.
14837 These filters accept the following parameters:
14839 c0 set first pixel component expression
14841 c1 set second pixel component expression
14843 c2 set third pixel component expression
14845 c3 set fourth pixel component expression, corresponds to the alpha
14847 component
14848 r set red component expression
14850 g set green component expression
14852 b set blue component expression
14854 a alpha component expression
14856 y set Y/luminance component expression
14858 u set U/Cb component expression
14860 v set V/Cr component expression
14862 Each of them specifies the expression to use for computing the lookup
14864 table for the corresponding pixel component values.
14865 The exact component associated to each of the c* options depends on the
14867 format in input.
14868 The lut filter requires either YUV or RGB pixel formats in input,
14870 lutrgb requires RGB pixel formats in input, and lutyuv requires YUV.
14871 The expressions can contain the following constants and functions:
14873 w
14875 h The input width and height.
14876 val The input value for the pixel component.
14878 clipval
14880 The input value, clipped to the minval-maxval range.
14881 maxval
14883 The maximum value for the pixel component.
14884 minval
14886 The minimum value for the pixel component.
14887 negval
14889 The negated value for the pixel component value, clipped to the
14890 minval-maxval range; it corresponds to the expression
14891 "maxval-clipval+minval".
14892 clip(val)
14894 The computed value in val, clipped to the minval-maxval range.
14895 gammaval(gamma)
14897 The computed gamma correction value of the pixel component value,
14898 clipped to the minval-maxval range. It corresponds to the
14899 expression
14900 "pow((clipval-minval)/(maxval-minval)\,gamma)*(maxval-minval)+minval"
14901 All expressions default to "clipval".
14903 Commands
14905 This filter supports same commands as options.
14907 Examples
14909 • Negate input video:
14911 lutrgb="r=maxval+minval-val:g=maxval+minval-val:b=maxval+minval-val"
14913 lutyuv="y=maxval+minval-val:u=maxval+minval-val:v=maxval+minval-val"
14914 The above is the same as:
14916 lutrgb="r=negval:g=negval:b=negval"
14918 lutyuv="y=negval:u=negval:v=negval"
14919 • Negate luminance:
14921 lutyuv=y=negval
14923 • Remove chroma components, turning the video into a graytone image:
14925 lutyuv="u=128:v=128"
14927 • Apply a luma burning effect:
14929 lutyuv="y=2*val"
14931 • Remove green and blue components:
14933 lutrgb="g=0:b=0"
14935 • Set a constant alpha channel value on input:
14937 format=rgba,lutrgb=a="maxval-minval/2"
14939 • Correct luminance gamma by a factor of 0.5:
14941 lutyuv=y=gammaval(0.5)
14943 • Discard least significant bits of luma:
14945 lutyuv=y='bitand(val, 128+64+32)'
14947 • Technicolor like effect:
14949 lutyuv=u='(val-maxval/2)*2+maxval/2':v='(val-maxval/2)*2+maxval/2'
14951 lut2, tlut2
14953 The "lut2" filter takes two input streams and outputs one stream.
14954 The "tlut2" (time lut2) filter takes two consecutive frames from one
14956 single stream.
14957 This filter accepts the following parameters:
14959 c0 set first pixel component expression
14961 c1 set second pixel component expression
14963 c2 set third pixel component expression
14965 c3 set fourth pixel component expression, corresponds to the alpha
14967 component
14968 d set output bit depth, only available for "lut2" filter. By default
14970 is 0, which means bit depth is automatically picked from first
14971 input format.
14972 The "lut2" filter also supports the framesync options.
14974 Each of them specifies the expression to use for computing the lookup
14976 table for the corresponding pixel component values.
14977 The exact component associated to each of the c* options depends on the
14979 format in inputs.
14980 The expressions can contain the following constants:
14982 w
14984 h The input width and height.
14985 x The first input value for the pixel component.
14987 y The second input value for the pixel component.
14989 bdx The first input video bit depth.
14991 bdy The second input video bit depth.
14993 All expressions default to "x".
14995 Commands
14997 This filter supports the all above options as commands except option
14999 "d".
15000 Examples
15002 • Highlight differences between two RGB video streams:
15004 lut2='ifnot(x-y,0,pow(2,bdx)-1):ifnot(x-y,0,pow(2,bdx)-1):ifnot(x-y,0,pow(2,bdx)-1)'
15006 • Highlight differences between two YUV video streams:
15008 lut2='ifnot(x-y,0,pow(2,bdx)-1):ifnot(x-y,pow(2,bdx-1),pow(2,bdx)-1):ifnot(x-y,pow(2,bdx-1),pow(2,bdx)-1)'
15010 • Show max difference between two video streams:
15012 lut2='if(lt(x,y),0,if(gt(x,y),pow(2,bdx)-1,pow(2,bdx-1))):if(lt(x,y),0,if(gt(x,y),pow(2,bdx)-1,pow(2,bdx-1))):if(lt(x,y),0,if(gt(x,y),pow(2,bdx)-1,pow(2,bdx-1)))'
15014 maskedclamp
15016 Clamp the first input stream with the second input and third input
15017 stream.
15018 Returns the value of first stream to be between second input stream -
15020 "undershoot" and third input stream + "overshoot".
15021 This filter accepts the following options:
15023 undershoot
15025 Default value is 0.
15026 overshoot
15028 Default value is 0.
15029 planes
15031 Set which planes will be processed as bitmap, unprocessed planes
15032 will be copied from first stream. By default value 0xf, all planes
15033 will be processed.
15034 Commands
15036 This filter supports the all above options as commands.
15038 maskedmax
15040 Merge the second and third input stream into output stream using
15041 absolute differences between second input stream and first input stream
15042 and absolute difference between third input stream and first input
15043 stream. The picked value will be from second input stream if second
15044 absolute difference is greater than first one or from third input
15045 stream otherwise.
15046 This filter accepts the following options:
15048 planes
15050 Set which planes will be processed as bitmap, unprocessed planes
15051 will be copied from first stream. By default value 0xf, all planes
15052 will be processed.
15053 Commands
15055 This filter supports the all above options as commands.
15057 maskedmerge
15059 Merge the first input stream with the second input stream using per
15060 pixel weights in the third input stream.
15061 A value of 0 in the third stream pixel component means that pixel
15063 component from first stream is returned unchanged, while maximum value
15064 (eg. 255 for 8-bit videos) means that pixel component from second
15065 stream is returned unchanged. Intermediate values define the amount of
15066 merging between both input stream's pixel components.
15067 This filter accepts the following options:
15069 planes
15071 Set which planes will be processed as bitmap, unprocessed planes
15072 will be copied from first stream. By default value 0xf, all planes
15073 will be processed.
15074 Commands
15076 This filter supports the all above options as commands.
15078 maskedmin
15080 Merge the second and third input stream into output stream using
15081 absolute differences between second input stream and first input stream
15082 and absolute difference between third input stream and first input
15083 stream. The picked value will be from second input stream if second
15084 absolute difference is less than first one or from third input stream
15085 otherwise.
15086 This filter accepts the following options:
15088 planes
15090 Set which planes will be processed as bitmap, unprocessed planes
15091 will be copied from first stream. By default value 0xf, all planes
15092 will be processed.
15093 Commands
15095 This filter supports the all above options as commands.
15097 maskedthreshold
15099 Pick pixels comparing absolute difference of two video streams with
15100 fixed threshold.
15101 If absolute difference between pixel component of first and second
15103 video stream is equal or lower than user supplied threshold than pixel
15104 component from first video stream is picked, otherwise pixel component
15105 from second video stream is picked.
15106 This filter accepts the following options:
15108 threshold
15110 Set threshold used when picking pixels from absolute difference
15111 from two input video streams.
15112 planes
15114 Set which planes will be processed as bitmap, unprocessed planes
15115 will be copied from second stream. By default value 0xf, all
15116 planes will be processed.
15117 mode
15119 Set mode of filter operation. Can be "abs" or "diff". Default is
15120 "abs".
15121 Commands
15123 This filter supports the all above options as commands.
15125 maskfun
15127 Create mask from input video.
15128 For example it is useful to create motion masks after "tblend" filter.
15130 This filter accepts the following options:
15132 low Set low threshold. Any pixel component lower or exact than this
15134 value will be set to 0.
15135 high
15137 Set high threshold. Any pixel component higher than this value will
15138 be set to max value allowed for current pixel format.
15139 planes
15141 Set planes to filter, by default all available planes are filtered.
15142 fill
15144 Fill all frame pixels with this value.
15145 sum Set max average pixel value for frame. If sum of all pixel
15147 components is higher that this average, output frame will be
15148 completely filled with value set by fill option. Typically useful
15149 for scene changes when used in combination with "tblend" filter.
15150 Commands
15152 This filter supports the all above options as commands.
15154 mcdeint
15156 Apply motion-compensation deinterlacing.
15157 It needs one field per frame as input and must thus be used together
15159 with yadif=1/3 or equivalent.
15160 This filter is only available in ffmpeg version 4.4 or earlier.
15162 This filter accepts the following options:
15164 mode
15166 Set the deinterlacing mode.
15167 It accepts one of the following values:
15169 fast
15171 medium
15172 slow
15173 use iterative motion estimation
15174 extra_slow
15176 like slow, but use multiple reference frames.
15177 Default value is fast.
15179 parity
15181 Set the picture field parity assumed for the input video. It must
15182 be one of the following values:
15183 0, tff
15185 assume top field first
15186 1, bff
15188 assume bottom field first
15189 Default value is bff.
15191 qp Set per-block quantization parameter (QP) used by the internal
15193 encoder.
15194 Higher values should result in a smoother motion vector field but
15196 less optimal individual vectors. Default value is 1.
15197 median
15199 Pick median pixel from certain rectangle defined by radius.
15200 This filter accepts the following options:
15202 radius
15204 Set horizontal radius size. Default value is 1. Allowed range is
15205 integer from 1 to 127.
15206 planes
15208 Set which planes to process. Default is 15, which is all available
15209 planes.
15210 radiusV
15212 Set vertical radius size. Default value is 0. Allowed range is
15213 integer from 0 to 127. If it is 0, value will be picked from
15214 horizontal "radius" option.
15215 percentile
15217 Set median percentile. Default value is 0.5. Default value of 0.5
15218 will pick always median values, while 0 will pick minimum values,
15219 and 1 maximum values.
15220 Commands
15222 This filter supports same commands as options. The command accepts the
15224 same syntax of the corresponding option.
15225 If the specified expression is not valid, it is kept at its current
15227 value.
15228 mergeplanes
15230 Merge color channel components from several video streams.
15231 The filter accepts up to 4 input streams, and merge selected input
15233 planes to the output video.
15234 This filter accepts the following options:
15236 mapping
15238 Set input to output plane mapping. Default is 0.
15239 The mappings is specified as a bitmap. It should be specified as a
15241 hexadecimal number in the form 0xAa[Bb[Cc[Dd]]]. 'Aa' describes the
15242 mapping for the first plane of the output stream. 'A' sets the
15243 number of the input stream to use (from 0 to 3), and 'a' the plane
15244 number of the corresponding input to use (from 0 to 3). The rest of
15245 the mappings is similar, 'Bb' describes the mapping for the output
15246 stream second plane, 'Cc' describes the mapping for the output
15247 stream third plane and 'Dd' describes the mapping for the output
15248 stream fourth plane.
15249 format
15251 Set output pixel format. Default is "yuva444p".
15252 map0s
15254 map1s
15255 map2s
15256 map3s
15257 Set input to output stream mapping for output Nth plane. Default is
15258 0.
15259 map0p
15261 map1p
15262 map2p
15263 map3p
15264 Set input to output plane mapping for output Nth plane. Default is
15265 0.
15266 Examples
15268 • Merge three gray video streams of same width and height into single
15270 video stream:
15271 [a0][a1][a2]mergeplanes=0x001020:yuv444p
15273 • Merge 1st yuv444p stream and 2nd gray video stream into yuva444p
15275 video stream:
15276 [a0][a1]mergeplanes=0x00010210:yuva444p
15278 • Swap Y and A plane in yuva444p stream:
15280 format=yuva444p,mergeplanes=0x03010200:yuva444p
15282 • Swap U and V plane in yuv420p stream:
15284 format=yuv420p,mergeplanes=0x000201:yuv420p
15286 • Cast a rgb24 clip to yuv444p:
15288 format=rgb24,mergeplanes=0x000102:yuv444p
15290 mestimate
15292 Estimate and export motion vectors using block matching algorithms.
15293 Motion vectors are stored in frame side data to be used by other
15294 filters.
15295 This filter accepts the following options:
15297 method
15299 Specify the motion estimation method. Accepts one of the following
15300 values:
15301 esa Exhaustive search algorithm.
15303 tss Three step search algorithm.
15305 tdls
15307 Two dimensional logarithmic search algorithm.
15308 ntss
15310 New three step search algorithm.
15311 fss Four step search algorithm.
15313 ds Diamond search algorithm.
15315 hexbs
15317 Hexagon-based search algorithm.
15318 epzs
15320 Enhanced predictive zonal search algorithm.
15321 umh Uneven multi-hexagon search algorithm.
15323 Default value is esa.
15325 mb_size
15327 Macroblock size. Default 16.
15328 search_param
15330 Search parameter. Default 7.
15331 midequalizer
15333 Apply Midway Image Equalization effect using two video streams.
15334 Midway Image Equalization adjusts a pair of images to have the same
15336 histogram, while maintaining their dynamics as much as possible. It's
15337 useful for e.g. matching exposures from a pair of stereo cameras.
15338 This filter has two inputs and one output, which must be of same pixel
15340 format, but may be of different sizes. The output of filter is first
15341 input adjusted with midway histogram of both inputs.
15342 This filter accepts the following option:
15344 planes
15346 Set which planes to process. Default is 15, which is all available
15347 planes.
15348 minterpolate
15350 Convert the video to specified frame rate using motion interpolation.
15351 This filter accepts the following options:
15353 fps Specify the output frame rate. This can be rational e.g.
15355 "60000/1001". Frames are dropped if fps is lower than source fps.
15356 Default 60.
15357 mi_mode
15359 Motion interpolation mode. Following values are accepted:
15360 dup Duplicate previous or next frame for interpolating new ones.
15362 blend
15364 Blend source frames. Interpolated frame is mean of previous and
15365 next frames.
15366 mci Motion compensated interpolation. Following options are
15368 effective when this mode is selected:
15369 mc_mode
15371 Motion compensation mode. Following values are accepted:
15372 obmc
15374 Overlapped block motion compensation.
15375 aobmc
15377 Adaptive overlapped block motion compensation. Window
15378 weighting coefficients are controlled adaptively
15379 according to the reliabilities of the neighboring
15380 motion vectors to reduce oversmoothing.
15381 Default mode is obmc.
15383 me_mode
15385 Motion estimation mode. Following values are accepted:
15386 bidir
15388 Bidirectional motion estimation. Motion vectors are
15389 estimated for each source frame in both forward and
15390 backward directions.
15391 bilat
15393 Bilateral motion estimation. Motion vectors are
15394 estimated directly for interpolated frame.
15395 Default mode is bilat.
15397 me The algorithm to be used for motion estimation. Following
15399 values are accepted:
15400 esa Exhaustive search algorithm.
15402 tss Three step search algorithm.
15404 tdls
15406 Two dimensional logarithmic search algorithm.
15407 ntss
15409 New three step search algorithm.
15410 fss Four step search algorithm.
15412 ds Diamond search algorithm.
15414 hexbs
15416 Hexagon-based search algorithm.
15417 epzs
15419 Enhanced predictive zonal search algorithm.
15420 umh Uneven multi-hexagon search algorithm.
15422 Default algorithm is epzs.
15424 mb_size
15426 Macroblock size. Default 16.
15427 search_param
15429 Motion estimation search parameter. Default 32.
15430 vsbmc
15432 Enable variable-size block motion compensation. Motion
15433 estimation is applied with smaller block sizes at object
15434 boundaries in order to make the them less blur. Default is
15435 0 (disabled).
15436 scd Scene change detection method. Scene change leads motion vectors to
15438 be in random direction. Scene change detection replace interpolated
15439 frames by duplicate ones. May not be needed for other modes.
15440 Following values are accepted:
15441 none
15443 Disable scene change detection.
15444 fdiff
15446 Frame difference. Corresponding pixel values are compared and
15447 if it satisfies scd_threshold scene change is detected.
15448 Default method is fdiff.
15450 scd_threshold
15452 Scene change detection threshold. Default is 10..
15453 mix
15455 Mix several video input streams into one video stream.
15456 A description of the accepted options follows.
15458 inputs
15460 The number of inputs. If unspecified, it defaults to 2.
15461 weights
15463 Specify weight of each input video stream as sequence. Each weight
15464 is separated by space. If number of weights is smaller than number
15465 of frames last specified weight will be used for all remaining
15466 unset weights.
15467 scale
15469 Specify scale, if it is set it will be multiplied with sum of each
15470 weight multiplied with pixel values to give final destination pixel
15471 value. By default scale is auto scaled to sum of weights.
15472 planes
15474 Set which planes to filter. Default is all. Allowed range is from 0
15475 to 15.
15476 duration
15478 Specify how end of stream is determined.
15479 longest
15481 The duration of the longest input. (default)
15482 shortest
15484 The duration of the shortest input.
15485 first
15487 The duration of the first input.
15488 Commands
15490 This filter supports the following commands:
15492 weights
15494 scale
15495 planes
15496 Syntax is same as option with same name.
15497 monochrome
15499 Convert video to gray using custom color filter.
15500 A description of the accepted options follows.
15502 cb Set the chroma blue spot. Allowed range is from -1 to 1. Default
15504 value is 0.
15505 cr Set the chroma red spot. Allowed range is from -1 to 1. Default
15507 value is 0.
15508 size
15510 Set the color filter size. Allowed range is from .1 to 10. Default
15511 value is 1.
15512 high
15514 Set the highlights strength. Allowed range is from 0 to 1. Default
15515 value is 0.
15516 Commands
15518 This filter supports the all above options as commands.
15520 morpho
15522 This filter allows to apply main morphological grayscale transforms,
15523 erode and dilate with arbitrary structures set in second input stream.
15524 Unlike naive implementation and much slower performance in erosion and
15526 dilation filters, when speed is critical "morpho" filter should be used
15527 instead.
15528 A description of accepted options follows,
15530 mode
15532 Set morphological transform to apply, can be:
15533 erode
15535 dilate
15536 open
15537 close
15538 gradient
15539 tophat
15540 blackhat
15541 Default is "erode".
15543 planes
15545 Set planes to filter, by default all planes except alpha are
15546 filtered.
15547 structure
15549 Set which structure video frames will be processed from second
15550 input stream, can be first or all. Default is all.
15551 The "morpho" filter also supports the framesync options.
15553 Commands
15555 This filter supports same commands as options.
15557 mpdecimate
15559 Drop frames that do not differ greatly from the previous frame in order
15560 to reduce frame rate.
15561 The main use of this filter is for very-low-bitrate encoding (e.g.
15563 streaming over dialup modem), but it could in theory be used for fixing
15564 movies that were inverse-telecined incorrectly.
15565 A description of the accepted options follows.
15567 max Set the maximum number of consecutive frames which can be dropped
15569 (if positive), or the minimum interval between dropped frames (if
15570 negative). If the value is 0, the frame is dropped disregarding the
15571 number of previous sequentially dropped frames.
15572 Default value is 0.
15574 hi
15576 lo
15577 frac
15578 Set the dropping threshold values.
15579 Values for hi and lo are for 8x8 pixel blocks and represent actual
15581 pixel value differences, so a threshold of 64 corresponds to 1 unit
15582 of difference for each pixel, or the same spread out differently
15583 over the block.
15584 A frame is a candidate for dropping if no 8x8 blocks differ by more
15586 than a threshold of hi, and if no more than frac blocks (1 meaning
15587 the whole image) differ by more than a threshold of lo.
15588 Default value for hi is 64*12, default value for lo is 64*5, and
15590 default value for frac is 0.33.
15591 msad
15593 Obtain the MSAD (Mean Sum of Absolute Differences) between two input
15594 videos.
15595 This filter takes two input videos.
15597 Both input videos must have the same resolution and pixel format for
15599 this filter to work correctly. Also it assumes that both inputs have
15600 the same number of frames, which are compared one by one.
15601 The obtained per component, average, min and max MSAD is printed
15603 through the logging system.
15604 The filter stores the calculated MSAD of each frame in frame metadata.
15606 This filter also supports the framesync options.
15608 In the below example the input file main.mpg being processed is
15610 compared with the reference file ref.mpg.
15611 ffmpeg -i main.mpg -i ref.mpg -lavfi msad -f null -
15613 multiply
15615 Multiply first video stream pixels values with second video stream
15616 pixels values.
15617 The filter accepts the following options:
15619 scale
15621 Set the scale applied to second video stream. By default is 1.
15622 Allowed range is from 0 to 9.
15623 offset
15625 Set the offset applied to second video stream. By default is 0.5.
15626 Allowed range is from -1 to 1.
15627 planes
15629 Specify planes from input video stream that will be processed. By
15630 default all planes are processed.
15631 Commands
15633 This filter supports same commands as options.
15635 negate
15637 Negate (invert) the input video.
15638 It accepts the following option:
15640 components
15642 Set components to negate.
15643 Available values for components are:
15645 y
15647 u
15648 v
15649 a
15650 r
15651 g
15652 b
15653 negate_alpha
15654 With value 1, it negates the alpha component, if present. Default
15655 value is 0.
15656 Commands
15658 This filter supports same commands as options.
15660 nlmeans
15662 Denoise frames using Non-Local Means algorithm.
15663 Each pixel is adjusted by looking for other pixels with similar
15665 contexts. This context similarity is defined by comparing their
15666 surrounding patches of size pxp. Patches are searched in an area of rxr
15667 around the pixel.
15668 Note that the research area defines centers for patches, which means
15670 some patches will be made of pixels outside that research area.
15671 The filter accepts the following options.
15673 s Set denoising strength. Default is 1.0. Must be in range [1.0,
15675 30.0].
15676 p Set patch size. Default is 7. Must be odd number in range [0, 99].
15678 pc Same as p but for chroma planes.
15680 The default value is 0 and means automatic.
15682 r Set research size. Default is 15. Must be odd number in range [0,
15684 99].
15685 rc Same as r but for chroma planes.
15687 The default value is 0 and means automatic.
15689 nnedi
15691 Deinterlace video using neural network edge directed interpolation.
15692 This filter accepts the following options:
15694 weights
15696 Mandatory option, without binary file filter can not work.
15697 Currently file can be found here:
15698 https://github.com/dubhater/vapoursynth-nnedi3/blob/master/src/nnedi3_weights.bin
15699 deint
15701 Set which frames to deinterlace, by default it is "all". Can be
15702 "all" or "interlaced".
15703 field
15705 Set mode of operation.
15706 Can be one of the following:
15708 af Use frame flags, both fields.
15710 a Use frame flags, single field.
15712 t Use top field only.
15714 b Use bottom field only.
15716 tf Use both fields, top first.
15718 bf Use both fields, bottom first.
15720 planes
15722 Set which planes to process, by default filter process all frames.
15723 nsize
15725 Set size of local neighborhood around each pixel, used by the
15726 predictor neural network.
15727 Can be one of the following:
15729 s8x6
15731 s16x6
15732 s32x6
15733 s48x6
15734 s8x4
15735 s16x4
15736 s32x4
15737 nns Set the number of neurons in predictor neural network. Can be one
15738 of the following:
15739 n16
15741 n32
15742 n64
15743 n128
15744 n256
15745 qual
15746 Controls the number of different neural network predictions that
15747 are blended together to compute the final output value. Can be
15748 "fast", default or "slow".
15749 etype
15751 Set which set of weights to use in the predictor. Can be one of
15752 the following:
15753 a, abs
15755 weights trained to minimize absolute error
15756 s, mse
15758 weights trained to minimize squared error
15759 pscrn
15761 Controls whether or not the prescreener neural network is used to
15762 decide which pixels should be processed by the predictor neural
15763 network and which can be handled by simple cubic interpolation.
15764 The prescreener is trained to know whether cubic interpolation will
15765 be sufficient for a pixel or whether it should be predicted by the
15766 predictor nn. The computational complexity of the prescreener nn
15767 is much less than that of the predictor nn. Since most pixels can
15768 be handled by cubic interpolation, using the prescreener generally
15769 results in much faster processing. The prescreener is pretty
15770 accurate, so the difference between using it and not using it is
15771 almost always unnoticeable.
15772 Can be one of the following:
15774 none
15776 original
15777 new
15778 new2
15779 new3
15780 Default is "new".
15782 Commands
15784 This filter supports same commands as options, excluding weights
15786 option.
15787 noformat
15789 Force libavfilter not to use any of the specified pixel formats for the
15790 input to the next filter.
15791 It accepts the following parameters:
15793 pix_fmts
15795 A '|'-separated list of pixel format names, such as
15796 pix_fmts=yuv420p|monow|rgb24".
15797 Examples
15799 • Force libavfilter to use a format different from yuv420p for the
15801 input to the vflip filter:
15802 noformat=pix_fmts=yuv420p,vflip
15804 • Convert the input video to any of the formats not contained in the
15806 list:
15807 noformat=yuv420p|yuv444p|yuv410p
15809 noise
15811 Add noise on video input frame.
15812 The filter accepts the following options:
15814 all_seed
15816 c0_seed
15817 c1_seed
15818 c2_seed
15819 c3_seed
15820 Set noise seed for specific pixel component or all pixel components
15821 in case of all_seed. Default value is 123457.
15822 all_strength, alls
15824 c0_strength, c0s
15825 c1_strength, c1s
15826 c2_strength, c2s
15827 c3_strength, c3s
15828 Set noise strength for specific pixel component or all pixel
15829 components in case all_strength. Default value is 0. Allowed range
15830 is [0, 100].
15831 all_flags, allf
15833 c0_flags, c0f
15834 c1_flags, c1f
15835 c2_flags, c2f
15836 c3_flags, c3f
15837 Set pixel component flags or set flags for all components if
15838 all_flags. Available values for component flags are:
15839 a averaged temporal noise (smoother)
15841 p mix random noise with a (semi)regular pattern
15843 t temporal noise (noise pattern changes between frames)
15845 u uniform noise (gaussian otherwise)
15847 Examples
15849 Add temporal and uniform noise to input video:
15851 noise=alls=20:allf=t+u
15853 normalize
15855 Normalize RGB video (aka histogram stretching, contrast stretching).
15856 See: https://en.wikipedia.org/wiki/Normalization_(image_processing)
15857 For each channel of each frame, the filter computes the input range and
15859 maps it linearly to the user-specified output range. The output range
15860 defaults to the full dynamic range from pure black to pure white.
15861 Temporal smoothing can be used on the input range to reduce flickering
15863 (rapid changes in brightness) caused when small dark or bright objects
15864 enter or leave the scene. This is similar to the auto-exposure
15865 (automatic gain control) on a video camera, and, like a video camera,
15866 it may cause a period of over- or under-exposure of the video.
15867 The R,G,B channels can be normalized independently, which may cause
15869 some color shifting, or linked together as a single channel, which
15870 prevents color shifting. Linked normalization preserves hue.
15871 Independent normalization does not, so it can be used to remove some
15872 color casts. Independent and linked normalization can be combined in
15873 any ratio.
15874 The normalize filter accepts the following options:
15876 blackpt
15878 whitept
15879 Colors which define the output range. The minimum input value is
15880 mapped to the blackpt. The maximum input value is mapped to the
15881 whitept. The defaults are black and white respectively. Specifying
15882 white for blackpt and black for whitept will give color-inverted,
15883 normalized video. Shades of grey can be used to reduce the dynamic
15884 range (contrast). Specifying saturated colors here can create some
15885 interesting effects.
15886 smoothing
15888 The number of previous frames to use for temporal smoothing. The
15889 input range of each channel is smoothed using a rolling average
15890 over the current frame and the smoothing previous frames. The
15891 default is 0 (no temporal smoothing).
15892 independence
15894 Controls the ratio of independent (color shifting) channel
15895 normalization to linked (color preserving) normalization. 0.0 is
15896 fully linked, 1.0 is fully independent. Defaults to 1.0 (fully
15897 independent).
15898 strength
15900 Overall strength of the filter. 1.0 is full strength. 0.0 is a
15901 rather expensive no-op. Defaults to 1.0 (full strength).
15902 Commands
15904 This filter supports same commands as options, excluding smoothing
15906 option. The command accepts the same syntax of the corresponding
15907 option.
15908 If the specified expression is not valid, it is kept at its current
15910 value.
15911 Examples
15913 Stretch video contrast to use the full dynamic range, with no temporal
15915 smoothing; may flicker depending on the source content:
15916 normalize=blackpt=black:whitept=white:smoothing=0
15918 As above, but with 50 frames of temporal smoothing; flicker should be
15920 reduced, depending on the source content:
15921 normalize=blackpt=black:whitept=white:smoothing=50
15923 As above, but with hue-preserving linked channel normalization:
15925 normalize=blackpt=black:whitept=white:smoothing=50:independence=0
15927 As above, but with half strength:
15929 normalize=blackpt=black:whitept=white:smoothing=50:independence=0:strength=0.5
15931 Map the darkest input color to red, the brightest input color to cyan:
15933 normalize=blackpt=red:whitept=cyan
15935 null
15937 Pass the video source unchanged to the output.
15938 ocr
15940 Optical Character Recognition
15941 This filter uses Tesseract for optical character recognition. To enable
15943 compilation of this filter, you need to configure FFmpeg with
15944 "--enable-libtesseract".
15945 It accepts the following options:
15947 datapath
15949 Set datapath to tesseract data. Default is to use whatever was set
15950 at installation.
15951 language
15953 Set language, default is "eng".
15954 whitelist
15956 Set character whitelist.
15957 blacklist
15959 Set character blacklist.
15960 The filter exports recognized text as the frame metadata
15962 "lavfi.ocr.text". The filter exports confidence of recognized words as
15963 the frame metadata "lavfi.ocr.confidence".
15964 ocv
15966 Apply a video transform using libopencv.
15967 To enable this filter, install the libopencv library and headers and
15969 configure FFmpeg with "--enable-libopencv".
15970 It accepts the following parameters:
15972 filter_name
15974 The name of the libopencv filter to apply.
15975 filter_params
15977 The parameters to pass to the libopencv filter. If not specified,
15978 the default values are assumed.
15979 Refer to the official libopencv documentation for more precise
15981 information:
15982 <http://docs.opencv.org/master/modules/imgproc/doc/filtering.html>
15983 Several libopencv filters are supported; see the following subsections.
15985 dilate
15987 Dilate an image by using a specific structuring element. It
15989 corresponds to the libopencv function "cvDilate".
15990 It accepts the parameters: struct_el|nb_iterations.
15992 struct_el represents a structuring element, and has the syntax:
15994 colsxrows+anchor_xxanchor_y/shape
15995 cols and rows represent the number of columns and rows of the
15997 structuring element, anchor_x and anchor_y the anchor point, and shape
15998 the shape for the structuring element. shape must be "rect", "cross",
15999 "ellipse", or "custom".
16000 If the value for shape is "custom", it must be followed by a string of
16002 the form "=filename". The file with name filename is assumed to
16003 represent a binary image, with each printable character corresponding
16004 to a bright pixel. When a custom shape is used, cols and rows are
16005 ignored, the number or columns and rows of the read file are assumed
16006 instead.
16007 The default value for struct_el is "3x3+0x0/rect".
16009 nb_iterations specifies the number of times the transform is applied to
16011 the image, and defaults to 1.
16012 Some examples:
16014 # Use the default values
16016 ocv=dilate
16017 # Dilate using a structuring element with a 5x5 cross, iterating two times
16019 ocv=filter_name=dilate:filter_params=5x5+2x2/cross|2
16020 # Read the shape from the file diamond.shape, iterating two times.
16022 # The file diamond.shape may contain a pattern of characters like this
16023 # *
16024 # ***
16025 # *****
16026 # ***
16027 # *
16028 # The specified columns and rows are ignored
16029 # but the anchor point coordinates are not
16030 ocv=dilate:0x0+2x2/custom=diamond.shape|2
16031 erode
16033 Erode an image by using a specific structuring element. It corresponds
16035 to the libopencv function "cvErode".
16036 It accepts the parameters: struct_el:nb_iterations, with the same
16038 syntax and semantics as the dilate filter.
16039 smooth
16041 Smooth the input video.
16043 The filter takes the following parameters:
16045 type|param1|param2|param3|param4.
16046 type is the type of smooth filter to apply, and must be one of the
16048 following values: "blur", "blur_no_scale", "median", "gaussian", or
16049 "bilateral". The default value is "gaussian".
16050 The meaning of param1, param2, param3, and param4 depends on the smooth
16052 type. param1 and param2 accept integer positive values or 0. param3 and
16053 param4 accept floating point values.
16054 The default value for param1 is 3. The default value for the other
16056 parameters is 0.
16057 These parameters correspond to the parameters assigned to the libopencv
16059 function "cvSmooth".
16060 oscilloscope
16062 2D Video Oscilloscope.
16063 Useful to measure spatial impulse, step responses, chroma delays, etc.
16065 It accepts the following parameters:
16067 x Set scope center x position.
16069 y Set scope center y position.
16071 s Set scope size, relative to frame diagonal.
16073 t Set scope tilt/rotation.
16075 o Set trace opacity.
16077 tx Set trace center x position.
16079 ty Set trace center y position.
16081 tw Set trace width, relative to width of frame.
16083 th Set trace height, relative to height of frame.
16085 c Set which components to trace. By default it traces first three
16087 components.
16088 g Draw trace grid. By default is enabled.
16090 st Draw some statistics. By default is enabled.
16092 sc Draw scope. By default is enabled.
16094 Commands
16096 This filter supports same commands as options. The command accepts the
16098 same syntax of the corresponding option.
16099 If the specified expression is not valid, it is kept at its current
16101 value.
16102 Examples
16104 • Inspect full first row of video frame.
16106 oscilloscope=x=0.5:y=0:s=1
16108 • Inspect full last row of video frame.
16110 oscilloscope=x=0.5:y=1:s=1
16112 • Inspect full 5th line of video frame of height 1080.
16114 oscilloscope=x=0.5:y=5/1080:s=1
16116 • Inspect full last column of video frame.
16118 oscilloscope=x=1:y=0.5:s=1:t=1
16120 overlay
16122 Overlay one video on top of another.
16123 It takes two inputs and has one output. The first input is the "main"
16125 video on which the second input is overlaid.
16126 It accepts the following parameters:
16128 A description of the accepted options follows.
16130 x
16132 y Set the expression for the x and y coordinates of the overlaid
16133 video on the main video. Default value is "0" for both expressions.
16134 In case the expression is invalid, it is set to a huge value
16135 (meaning that the overlay will not be displayed within the output
16136 visible area).
16137 eof_action
16139 See framesync.
16140 eval
16142 Set when the expressions for x, and y are evaluated.
16143 It accepts the following values:
16145 init
16147 only evaluate expressions once during the filter initialization
16148 or when a command is processed
16149 frame
16151 evaluate expressions for each incoming frame
16152 Default value is frame.
16154 shortest
16156 See framesync.
16157 format
16159 Set the format for the output video.
16160 It accepts the following values:
16162 yuv420
16164 force YUV420 output
16165 yuv420p10
16167 force YUV420p10 output
16168 yuv422
16170 force YUV422 output
16171 yuv422p10
16173 force YUV422p10 output
16174 yuv444
16176 force YUV444 output
16177 rgb force packed RGB output
16179 gbrp
16181 force planar RGB output
16182 auto
16184 automatically pick format
16185 Default value is yuv420.
16187 repeatlast
16189 See framesync.
16190 alpha
16192 Set format of alpha of the overlaid video, it can be straight or
16193 premultiplied. Default is straight.
16194 The x, and y expressions can contain the following parameters.
16196 main_w, W
16198 main_h, H
16199 The main input width and height.
16200 overlay_w, w
16202 overlay_h, h
16203 The overlay input width and height.
16204 x
16206 y The computed values for x and y. They are evaluated for each new
16207 frame.
16208 hsub
16210 vsub
16211 horizontal and vertical chroma subsample values of the output
16212 format. For example for the pixel format "yuv422p" hsub is 2 and
16213 vsub is 1.
16214 n the number of input frame, starting from 0
16216 pos the position in the file of the input frame, NAN if unknown
16218 t The timestamp, expressed in seconds. It's NAN if the input
16220 timestamp is unknown.
16221 This filter also supports the framesync options.
16223 Note that the n, pos, t variables are available only when evaluation is
16225 done per frame, and will evaluate to NAN when eval is set to init.
16226 Be aware that frames are taken from each input video in timestamp
16228 order, hence, if their initial timestamps differ, it is a good idea to
16229 pass the two inputs through a setpts=PTS-STARTPTS filter to have them
16230 begin in the same zero timestamp, as the example for the movie filter
16231 does.
16232 You can chain together more overlays but you should test the efficiency
16234 of such approach.
16235 Commands
16237 This filter supports the following commands:
16239 x
16241 y Modify the x and y of the overlay input. The command accepts the
16242 same syntax of the corresponding option.
16243 If the specified expression is not valid, it is kept at its current
16245 value.
16246 Examples
16248 • Draw the overlay at 10 pixels from the bottom right corner of the
16250 main video:
16251 overlay=main_w-overlay_w-10:main_h-overlay_h-10
16253 Using named options the example above becomes:
16255 overlay=x=main_w-overlay_w-10:y=main_h-overlay_h-10
16257 • Insert a transparent PNG logo in the bottom left corner of the
16259 input, using the ffmpeg tool with the "-filter_complex" option:
16260 ffmpeg -i input -i logo -filter_complex 'overlay=10:main_h-overlay_h-10' output
16262 • Insert 2 different transparent PNG logos (second logo on bottom
16264 right corner) using the ffmpeg tool:
16265 ffmpeg -i input -i logo1 -i logo2 -filter_complex 'overlay=x=10:y=H-h-10,overlay=x=W-w-10:y=H-h-10' output
16267 • Add a transparent color layer on top of the main video; "WxH" must
16269 specify the size of the main input to the overlay filter:
16270 color=color=red@.3:size=WxH [over]; [in][over] overlay [out]
16272 • Play an original video and a filtered version (here with the
16274 deshake filter) side by side using the ffplay tool:
16275 ffplay input.avi -vf 'split[a][b]; [a]pad=iw*2:ih[src]; [b]deshake[filt]; [src][filt]overlay=w'
16277 The above command is the same as:
16279 ffplay input.avi -vf 'split[b], pad=iw*2[src], [b]deshake, [src]overlay=w'
16281 • Make a sliding overlay appearing from the left to the right top
16283 part of the screen starting since time 2:
16284 overlay=x='if(gte(t,2), -w+(t-2)*20, NAN)':y=0
16286 • Compose output by putting two input videos side to side:
16288 ffmpeg -i left.avi -i right.avi -filter_complex "
16290 nullsrc=size=200x100 [background];
16291 [0:v] setpts=PTS-STARTPTS, scale=100x100 [left];
16292 [1:v] setpts=PTS-STARTPTS, scale=100x100 [right];
16293 [background][left] overlay=shortest=1 [background+left];
16294 [background+left][right] overlay=shortest=1:x=100 [left+right]
16295 "
16296 • Mask 10-20 seconds of a video by applying the delogo filter to a
16298 section
16299 ffmpeg -i test.avi -codec:v:0 wmv2 -ar 11025 -b:v 9000k
16301 -vf '[in]split[split_main][split_delogo];[split_delogo]trim=start=360:end=371,delogo=0:0:640:480[delogoed];[split_main][delogoed]overlay=eof_action=pass[out]'
16302 masked.avi
16303 • Chain several overlays in cascade:
16305 nullsrc=s=200x200 [bg];
16307 testsrc=s=100x100, split=4 [in0][in1][in2][in3];
16308 [in0] lutrgb=r=0, [bg] overlay=0:0 [mid0];
16309 [in1] lutrgb=g=0, [mid0] overlay=100:0 [mid1];
16310 [in2] lutrgb=b=0, [mid1] overlay=0:100 [mid2];
16311 [in3] null, [mid2] overlay=100:100 [out0]
16312 overlay_cuda
16314 Overlay one video on top of another.
16315 This is the CUDA variant of the overlay filter. It only accepts CUDA
16317 frames. The underlying input pixel formats have to match.
16318 It takes two inputs and has one output. The first input is the "main"
16320 video on which the second input is overlaid.
16321 It accepts the following parameters:
16323 x
16325 y Set expressions for the x and y coordinates of the overlaid video
16326 on the main video.
16327 They can contain the following parameters:
16329 main_w, W
16331 main_h, H
16332 The main input width and height.
16333 overlay_w, w
16335 overlay_h, h
16336 The overlay input width and height.
16337 x
16339 y The computed values for x and y. They are evaluated for each
16340 new frame.
16341 n The ordinal index of the main input frame, starting from 0.
16343 pos The byte offset position in the file of the main input frame,
16345 NAN if unknown.
16346 t The timestamp of the main input frame, expressed in seconds,
16348 NAN if unknown.
16349 Default value is "0" for both expressions.
16351 eval
16353 Set when the expressions for x and y are evaluated.
16354 It accepts the following values:
16356 init
16358 Evaluate expressions once during filter initialization or when
16359 a command is processed.
16360 frame
16362 Evaluate expressions for each incoming frame
16363 Default value is frame.
16365 eof_action
16367 See framesync.
16368 shortest
16370 See framesync.
16371 repeatlast
16373 See framesync.
16374 This filter also supports the framesync options.
16376 owdenoise
16378 Apply Overcomplete Wavelet denoiser.
16379 The filter accepts the following options:
16381 depth
16383 Set depth.
16384 Larger depth values will denoise lower frequency components more,
16386 but slow down filtering.
16387 Must be an int in the range 8-16, default is 8.
16389 luma_strength, ls
16391 Set luma strength.
16392 Must be a double value in the range 0-1000, default is 1.0.
16394 chroma_strength, cs
16396 Set chroma strength.
16397 Must be a double value in the range 0-1000, default is 1.0.
16399 pad
16401 Add paddings to the input image, and place the original input at the
16402 provided x, y coordinates.
16403 It accepts the following parameters:
16405 width, w
16407 height, h
16408 Specify an expression for the size of the output image with the
16409 paddings added. If the value for width or height is 0, the
16410 corresponding input size is used for the output.
16411 The width expression can reference the value set by the height
16413 expression, and vice versa.
16414 The default value of width and height is 0.
16416 x
16418 y Specify the offsets to place the input image at within the padded
16419 area, with respect to the top/left border of the output image.
16420 The x expression can reference the value set by the y expression,
16422 and vice versa.
16423 The default value of x and y is 0.
16425 If x or y evaluate to a negative number, they'll be changed so the
16427 input image is centered on the padded area.
16428 color
16430 Specify the color of the padded area. For the syntax of this
16431 option, check the "Color" section in the ffmpeg-utils manual.
16432 The default value of color is "black".
16434 eval
16436 Specify when to evaluate width, height, x and y expression.
16437 It accepts the following values:
16439 init
16441 Only evaluate expressions once during the filter initialization
16442 or when a command is processed.
16443 frame
16445 Evaluate expressions for each incoming frame.
16446 Default value is init.
16448 aspect
16450 Pad to aspect instead to a resolution.
16451 The value for the width, height, x, and y options are expressions
16453 containing the following constants:
16454 in_w
16456 in_h
16457 The input video width and height.
16458 iw
16460 ih These are the same as in_w and in_h.
16461 out_w
16463 out_h
16464 The output width and height (the size of the padded area), as
16465 specified by the width and height expressions.
16466 ow
16468 oh These are the same as out_w and out_h.
16469 x
16471 y The x and y offsets as specified by the x and y expressions, or NAN
16472 if not yet specified.
16473 a same as iw / ih
16475 sar input sample aspect ratio
16477 dar input display aspect ratio, it is the same as (iw / ih) * sar
16479 hsub
16481 vsub
16482 The horizontal and vertical chroma subsample values. For example
16483 for the pixel format "yuv422p" hsub is 2 and vsub is 1.
16484 Examples
16486 • Add paddings with the color "violet" to the input video. The output
16488 video size is 640x480, and the top-left corner of the input video
16489 is placed at column 0, row 40
16490 pad=640:480:0:40:violet
16492 The example above is equivalent to the following command:
16494 pad=width=640:height=480:x=0:y=40:color=violet
16496 • Pad the input to get an output with dimensions increased by 3/2,
16498 and put the input video at the center of the padded area:
16499 pad="3/2*iw:3/2*ih:(ow-iw)/2:(oh-ih)/2"
16501 • Pad the input to get a squared output with size equal to the
16503 maximum value between the input width and height, and put the input
16504 video at the center of the padded area:
16505 pad="max(iw\,ih):ow:(ow-iw)/2:(oh-ih)/2"
16507 • Pad the input to get a final w/h ratio of 16:9:
16509 pad="ih*16/9:ih:(ow-iw)/2:(oh-ih)/2"
16511 • In case of anamorphic video, in order to set the output display
16513 aspect correctly, it is necessary to use sar in the expression,
16514 according to the relation:
16515 (ih * X / ih) * sar = output_dar
16517 X = output_dar / sar
16518 Thus the previous example needs to be modified to:
16520 pad="ih*16/9/sar:ih:(ow-iw)/2:(oh-ih)/2"
16522 • Double the output size and put the input video in the bottom-right
16524 corner of the output padded area:
16525 pad="2*iw:2*ih:ow-iw:oh-ih"
16527 palettegen
16529 Generate one palette for a whole video stream.
16530 It accepts the following options:
16532 max_colors
16534 Set the maximum number of colors to quantize in the palette. Note:
16535 the palette will still contain 256 colors; the unused palette
16536 entries will be black.
16537 reserve_transparent
16539 Create a palette of 255 colors maximum and reserve the last one for
16540 transparency. Reserving the transparency color is useful for GIF
16541 optimization. If not set, the maximum of colors in the palette
16542 will be 256. You probably want to disable this option for a
16543 standalone image. Set by default.
16544 transparency_color
16546 Set the color that will be used as background for transparency.
16547 stats_mode
16549 Set statistics mode.
16550 It accepts the following values:
16552 full
16554 Compute full frame histograms.
16555 diff
16557 Compute histograms only for the part that differs from previous
16558 frame. This might be relevant to give more importance to the
16559 moving part of your input if the background is static.
16560 single
16562 Compute new histogram for each frame.
16563 Default value is full.
16565 The filter also exports the frame metadata "lavfi.color_quant_ratio"
16567 ("nb_color_in / nb_color_out") which you can use to evaluate the degree
16568 of color quantization of the palette. This information is also visible
16569 at info logging level.
16570 Examples
16572 • Generate a representative palette of a given video using ffmpeg:
16574 ffmpeg -i input.mkv -vf palettegen palette.png
16576 paletteuse
16578 Use a palette to downsample an input video stream.
16579 The filter takes two inputs: one video stream and a palette. The
16581 palette must be a 256 pixels image.
16582 It accepts the following options:
16584 dither
16586 Select dithering mode. Available algorithms are:
16587 bayer
16589 Ordered 8x8 bayer dithering (deterministic)
16590 heckbert
16592 Dithering as defined by Paul Heckbert in 1982 (simple error
16593 diffusion). Note: this dithering is sometimes considered
16594 "wrong" and is included as a reference.
16595 floyd_steinberg
16597 Floyd and Steingberg dithering (error diffusion)
16598 sierra2
16600 Frankie Sierra dithering v2 (error diffusion)
16601 sierra2_4a
16603 Frankie Sierra dithering v2 "Lite" (error diffusion)
16604 sierra3
16606 Frankie Sierra dithering v3 (error diffusion)
16607 burkes
16609 Burkes dithering (error diffusion)
16610 atkinson
16612 Atkinson dithering by Bill Atkinson at Apple Computer (error
16613 diffusion)
16614 Default is sierra2_4a.
16616 bayer_scale
16618 When bayer dithering is selected, this option defines the scale of
16619 the pattern (how much the crosshatch pattern is visible). A low
16620 value means more visible pattern for less banding, and higher value
16621 means less visible pattern at the cost of more banding.
16622 The option must be an integer value in the range [0,5]. Default is
16624 2.
16625 diff_mode
16627 If set, define the zone to process
16628 rectangle
16630 Only the changing rectangle will be reprocessed. This is
16631 similar to GIF cropping/offsetting compression mechanism. This
16632 option can be useful for speed if only a part of the image is
16633 changing, and has use cases such as limiting the scope of the
16634 error diffusal dither to the rectangle that bounds the moving
16635 scene (it leads to more deterministic output if the scene
16636 doesn't change much, and as a result less moving noise and
16637 better GIF compression).
16638 Default is none.
16640 new Take new palette for each output frame.
16642 alpha_threshold
16644 Sets the alpha threshold for transparency. Alpha values above this
16645 threshold will be treated as completely opaque, and values below
16646 this threshold will be treated as completely transparent.
16647 The option must be an integer value in the range [0,255]. Default
16649 is 128.
16650 Examples
16652 • Use a palette (generated for example with palettegen) to encode a
16654 GIF using ffmpeg:
16655 ffmpeg -i input.mkv -i palette.png -lavfi paletteuse output.gif
16657 perspective
16659 Correct perspective of video not recorded perpendicular to the screen.
16660 A description of the accepted parameters follows.
16662 x0
16664 y0
16665 x1
16666 y1
16667 x2
16668 y2
16669 x3
16670 y3 Set coordinates expression for top left, top right, bottom left and
16671 bottom right corners. Default values are "0:0:W:0:0:H:W:H" with
16672 which perspective will remain unchanged. If the "sense" option is
16673 set to "source", then the specified points will be sent to the
16674 corners of the destination. If the "sense" option is set to
16675 "destination", then the corners of the source will be sent to the
16676 specified coordinates.
16677 The expressions can use the following variables:
16679 W
16681 H the width and height of video frame.
16682 in Input frame count.
16684 on Output frame count.
16686 interpolation
16688 Set interpolation for perspective correction.
16689 It accepts the following values:
16691 linear
16693 cubic
16694 Default value is linear.
16696 sense
16698 Set interpretation of coordinate options.
16699 It accepts the following values:
16701 0, source
16703 Send point in the source specified by the given coordinates to
16704 the corners of the destination.
16705 1, destination
16707 Send the corners of the source to the point in the destination
16708 specified by the given coordinates.
16709 Default value is source.
16711 eval
16713 Set when the expressions for coordinates x0,y0,...x3,y3 are
16714 evaluated.
16715 It accepts the following values:
16717 init
16719 only evaluate expressions once during the filter initialization
16720 or when a command is processed
16721 frame
16723 evaluate expressions for each incoming frame
16724 Default value is init.
16726 phase
16728 Delay interlaced video by one field time so that the field order
16729 changes.
16730 The intended use is to fix PAL movies that have been captured with the
16732 opposite field order to the film-to-video transfer.
16733 A description of the accepted parameters follows.
16735 mode
16737 Set phase mode.
16738 It accepts the following values:
16740 t Capture field order top-first, transfer bottom-first. Filter
16742 will delay the bottom field.
16743 b Capture field order bottom-first, transfer top-first. Filter
16745 will delay the top field.
16746 p Capture and transfer with the same field order. This mode only
16748 exists for the documentation of the other options to refer to,
16749 but if you actually select it, the filter will faithfully do
16750 nothing.
16751 a Capture field order determined automatically by field flags,
16753 transfer opposite. Filter selects among t and b modes on a
16754 frame by frame basis using field flags. If no field information
16755 is available, then this works just like u.
16756 u Capture unknown or varying, transfer opposite. Filter selects
16758 among t and b on a frame by frame basis by analyzing the images
16759 and selecting the alternative that produces best match between
16760 the fields.
16761 T Capture top-first, transfer unknown or varying. Filter selects
16763 among t and p using image analysis.
16764 B Capture bottom-first, transfer unknown or varying. Filter
16766 selects among b and p using image analysis.
16767 A Capture determined by field flags, transfer unknown or varying.
16769 Filter selects among t, b and p using field flags and image
16770 analysis. If no field information is available, then this works
16771 just like U. This is the default mode.
16772 U Both capture and transfer unknown or varying. Filter selects
16774 among t, b and p using image analysis only.
16775 Commands
16777 This filter supports the all above options as commands.
16779 photosensitivity
16781 Reduce various flashes in video, so to help users with epilepsy.
16782 It accepts the following options:
16784 frames, f
16786 Set how many frames to use when filtering. Default is 30.
16787 threshold, t
16789 Set detection threshold factor. Default is 1. Lower is stricter.
16790 skip
16792 Set how many pixels to skip when sampling frames. Default is 1.
16793 Allowed range is from 1 to 1024.
16794 bypass
16796 Leave frames unchanged. Default is disabled.
16797 pixdesctest
16799 Pixel format descriptor test filter, mainly useful for internal
16800 testing. The output video should be equal to the input video.
16801 For example:
16803 format=monow, pixdesctest
16805 can be used to test the monowhite pixel format descriptor definition.
16807 pixelize
16809 Apply pixelization to video stream.
16810 The filter accepts the following options:
16812 width, w
16814 height, h
16815 Set block dimensions that will be used for pixelization. Default
16816 value is 16.
16817 mode, m
16819 Set the mode of pixelization used.
16820 Possible values are:
16822 avg
16824 min
16825 max
16826 Default value is "avg".
16828 planes, p
16830 Set what planes to filter. Default is to filter all planes.
16831 Commands
16833 This filter supports all options as commands.
16835 pixscope
16837 Display sample values of color channels. Mainly useful for checking
16838 color and levels. Minimum supported resolution is 640x480.
16839 The filters accept the following options:
16841 x Set scope X position, relative offset on X axis.
16843 y Set scope Y position, relative offset on Y axis.
16845 w Set scope width.
16847 h Set scope height.
16849 o Set window opacity. This window also holds statistics about pixel
16851 area.
16852 wx Set window X position, relative offset on X axis.
16854 wy Set window Y position, relative offset on Y axis.
16856 Commands
16858 This filter supports same commands as options.
16860 pp
16862 Enable the specified chain of postprocessing subfilters using
16863 libpostproc. This library should be automatically selected with a GPL
16864 build ("--enable-gpl"). Subfilters must be separated by '/' and can be
16865 disabled by prepending a '-'. Each subfilter and some options have a
16866 short and a long name that can be used interchangeably, i.e. dr/dering
16867 are the same.
16868 The filters accept the following options:
16870 subfilters
16872 Set postprocessing subfilters string.
16873 All subfilters share common options to determine their scope:
16875 a/autoq
16877 Honor the quality commands for this subfilter.
16878 c/chrom
16880 Do chrominance filtering, too (default).
16881 y/nochrom
16883 Do luminance filtering only (no chrominance).
16884 n/noluma
16886 Do chrominance filtering only (no luminance).
16887 These options can be appended after the subfilter name, separated by a
16889 '|'.
16890 Available subfilters are:
16892 hb/hdeblock[|difference[|flatness]]
16894 Horizontal deblocking filter
16895 difference
16897 Difference factor where higher values mean more deblocking
16898 (default: 32).
16899 flatness
16901 Flatness threshold where lower values mean more deblocking
16902 (default: 39).
16903 vb/vdeblock[|difference[|flatness]]
16905 Vertical deblocking filter
16906 difference
16908 Difference factor where higher values mean more deblocking
16909 (default: 32).
16910 flatness
16912 Flatness threshold where lower values mean more deblocking
16913 (default: 39).
16914 ha/hadeblock[|difference[|flatness]]
16916 Accurate horizontal deblocking filter
16917 difference
16919 Difference factor where higher values mean more deblocking
16920 (default: 32).
16921 flatness
16923 Flatness threshold where lower values mean more deblocking
16924 (default: 39).
16925 va/vadeblock[|difference[|flatness]]
16927 Accurate vertical deblocking filter
16928 difference
16930 Difference factor where higher values mean more deblocking
16931 (default: 32).
16932 flatness
16934 Flatness threshold where lower values mean more deblocking
16935 (default: 39).
16936 The horizontal and vertical deblocking filters share the difference and
16938 flatness values so you cannot set different horizontal and vertical
16939 thresholds.
16940 h1/x1hdeblock
16942 Experimental horizontal deblocking filter
16943 v1/x1vdeblock
16945 Experimental vertical deblocking filter
16946 dr/dering
16948 Deringing filter
16949 tn/tmpnoise[|threshold1[|threshold2[|threshold3]]], temporal noise
16951 reducer
16952 threshold1
16953 larger -> stronger filtering
16954 threshold2
16956 larger -> stronger filtering
16957 threshold3
16959 larger -> stronger filtering
16960 al/autolevels[:f/fullyrange], automatic brightness / contrast
16962 correction
16963 f/fullyrange
16964 Stretch luminance to "0-255".
16965 lb/linblenddeint
16967 Linear blend deinterlacing filter that deinterlaces the given block
16968 by filtering all lines with a "(1 2 1)" filter.
16969 li/linipoldeint
16971 Linear interpolating deinterlacing filter that deinterlaces the
16972 given block by linearly interpolating every second line.
16973 ci/cubicipoldeint
16975 Cubic interpolating deinterlacing filter deinterlaces the given
16976 block by cubically interpolating every second line.
16977 md/mediandeint
16979 Median deinterlacing filter that deinterlaces the given block by
16980 applying a median filter to every second line.
16981 fd/ffmpegdeint
16983 FFmpeg deinterlacing filter that deinterlaces the given block by
16984 filtering every second line with a "(-1 4 2 4 -1)" filter.
16985 l5/lowpass5
16987 Vertically applied FIR lowpass deinterlacing filter that
16988 deinterlaces the given block by filtering all lines with a "(-1 2 6
16989 2 -1)" filter.
16990 fq/forceQuant[|quantizer]
16992 Overrides the quantizer table from the input with the constant
16993 quantizer you specify.
16994 quantizer
16996 Quantizer to use
16997 de/default
16999 Default pp filter combination ("hb|a,vb|a,dr|a")
17000 fa/fast
17002 Fast pp filter combination ("h1|a,v1|a,dr|a")
17003 ac High quality pp filter combination ("ha|a|128|7,va|a,dr|a")
17005 Examples
17007 • Apply horizontal and vertical deblocking, deringing and automatic
17009 brightness/contrast:
17010 pp=hb/vb/dr/al
17012 • Apply default filters without brightness/contrast correction:
17014 pp=de/-al
17016 • Apply default filters and temporal denoiser:
17018 pp=default/tmpnoise|1|2|3
17020 • Apply deblocking on luminance only, and switch vertical deblocking
17022 on or off automatically depending on available CPU time:
17023 pp=hb|y/vb|a
17025 pp7
17027 Apply Postprocessing filter 7. It is variant of the spp filter, similar
17028 to spp = 6 with 7 point DCT, where only the center sample is used after
17029 IDCT.
17030 The filter accepts the following options:
17032 qp Force a constant quantization parameter. It accepts an integer in
17034 range 0 to 63. If not set, the filter will use the QP from the
17035 video stream (if available).
17036 mode
17038 Set thresholding mode. Available modes are:
17039 hard
17041 Set hard thresholding.
17042 soft
17044 Set soft thresholding (better de-ringing effect, but likely
17045 blurrier).
17046 medium
17048 Set medium thresholding (good results, default).
17049 premultiply
17051 Apply alpha premultiply effect to input video stream using first plane
17052 of second stream as alpha.
17053 Both streams must have same dimensions and same pixel format.
17055 The filter accepts the following option:
17057 planes
17059 Set which planes will be processed, unprocessed planes will be
17060 copied. By default value 0xf, all planes will be processed.
17061 inplace
17063 Do not require 2nd input for processing, instead use alpha plane
17064 from input stream.
17065 prewitt
17067 Apply prewitt operator to input video stream.
17068 The filter accepts the following option:
17070 planes
17072 Set which planes will be processed, unprocessed planes will be
17073 copied. By default value 0xf, all planes will be processed.
17074 scale
17076 Set value which will be multiplied with filtered result.
17077 delta
17079 Set value which will be added to filtered result.
17080 Commands
17082 This filter supports the all above options as commands.
17084 pseudocolor
17086 Alter frame colors in video with pseudocolors.
17087 This filter accepts the following options:
17089 c0 set pixel first component expression
17091 c1 set pixel second component expression
17093 c2 set pixel third component expression
17095 c3 set pixel fourth component expression, corresponds to the alpha
17097 component
17098 index, i
17100 set component to use as base for altering colors
17101 preset, p
17103 Pick one of built-in LUTs. By default is set to none.
17104 Available LUTs:
17106 magma
17108 inferno
17109 plasma
17110 viridis
17111 turbo
17112 cividis
17113 range1
17114 range2
17115 shadows
17116 highlights
17117 solar
17118 nominal
17119 preferred
17120 total
17121 spectral
17122 opacity
17123 Set opacity of output colors. Allowed range is from 0 to 1.
17124 Default value is set to 1.
17125 Each of the expression options specifies the expression to use for
17127 computing the lookup table for the corresponding pixel component
17128 values.
17129 The expressions can contain the following constants and functions:
17131 w
17133 h The input width and height.
17134 val The input value for the pixel component.
17136 ymin, umin, vmin, amin
17138 The minimum allowed component value.
17139 ymax, umax, vmax, amax
17141 The maximum allowed component value.
17142 All expressions default to "val".
17144 Commands
17146 This filter supports the all above options as commands.
17148 Examples
17150 • Change too high luma values to gradient:
17152 pseudocolor="'if(between(val,ymax,amax),lerp(ymin,ymax,(val-ymax)/(amax-ymax)),-1):if(between(val,ymax,amax),lerp(umax,umin,(val-ymax)/(amax-ymax)),-1):if(between(val,ymax,amax),lerp(vmin,vmax,(val-ymax)/(amax-ymax)),-1):-1'"
17154 psnr
17156 Obtain the average, maximum and minimum PSNR (Peak Signal to Noise
17157 Ratio) between two input videos.
17158 This filter takes in input two input videos, the first input is
17160 considered the "main" source and is passed unchanged to the output. The
17161 second input is used as a "reference" video for computing the PSNR.
17162 Both video inputs must have the same resolution and pixel format for
17164 this filter to work correctly. Also it assumes that both inputs have
17165 the same number of frames, which are compared one by one.
17166 The obtained average PSNR is printed through the logging system.
17168 The filter stores the accumulated MSE (mean squared error) of each
17170 frame, and at the end of the processing it is averaged across all
17171 frames equally, and the following formula is applied to obtain the
17172 PSNR:
17173 PSNR = 10*log10(MAX^2/MSE)
17175 Where MAX is the average of the maximum values of each component of the
17177 image.
17178 The description of the accepted parameters follows.
17180 stats_file, f
17182 If specified the filter will use the named file to save the PSNR of
17183 each individual frame. When filename equals "-" the data is sent to
17184 standard output.
17185 stats_version
17187 Specifies which version of the stats file format to use. Details of
17188 each format are written below. Default value is 1.
17189 stats_add_max
17191 Determines whether the max value is output to the stats log.
17192 Default value is 0. Requires stats_version >= 2. If this is set
17193 and stats_version < 2, the filter will return an error.
17194 This filter also supports the framesync options.
17196 The file printed if stats_file is selected, contains a sequence of
17198 key/value pairs of the form key:value for each compared couple of
17199 frames.
17200 If a stats_version greater than 1 is specified, a header line precedes
17202 the list of per-frame-pair stats, with key value pairs following the
17203 frame format with the following parameters:
17204 psnr_log_version
17206 The version of the log file format. Will match stats_version.
17207 fields
17209 A comma separated list of the per-frame-pair parameters included in
17210 the log.
17211 A description of each shown per-frame-pair parameter follows:
17213 n sequential number of the input frame, starting from 1
17215 mse_avg
17217 Mean Square Error pixel-by-pixel average difference of the compared
17218 frames, averaged over all the image components.
17219 mse_y, mse_u, mse_v, mse_r, mse_g, mse_b, mse_a
17221 Mean Square Error pixel-by-pixel average difference of the compared
17222 frames for the component specified by the suffix.
17223 psnr_y, psnr_u, psnr_v, psnr_r, psnr_g, psnr_b, psnr_a
17225 Peak Signal to Noise ratio of the compared frames for the component
17226 specified by the suffix.
17227 max_avg, max_y, max_u, max_v
17229 Maximum allowed value for each channel, and average over all
17230 channels.
17231 Examples
17233 • For example:
17235 movie=ref_movie.mpg, setpts=PTS-STARTPTS [main];
17237 [main][ref] psnr="stats_file=stats.log" [out]
17238 On this example the input file being processed is compared with the
17240 reference file ref_movie.mpg. The PSNR of each individual frame is
17241 stored in stats.log.
17242 • Another example with different containers:
17244 ffmpeg -i main.mpg -i ref.mkv -lavfi "[0:v]settb=AVTB,setpts=PTS-STARTPTS[main];[1:v]settb=AVTB,setpts=PTS-STARTPTS[ref];[main][ref]psnr" -f null -
17246 pullup
17248 Pulldown reversal (inverse telecine) filter, capable of handling mixed
17249 hard-telecine, 24000/1001 fps progressive, and 30000/1001 fps
17250 progressive content.
17251 The pullup filter is designed to take advantage of future context in
17253 making its decisions. This filter is stateless in the sense that it
17254 does not lock onto a pattern to follow, but it instead looks forward to
17255 the following fields in order to identify matches and rebuild
17256 progressive frames.
17257 To produce content with an even framerate, insert the fps filter after
17259 pullup, use "fps=24000/1001" if the input frame rate is 29.97fps,
17260 "fps=24" for 30fps and the (rare) telecined 25fps input.
17261 The filter accepts the following options:
17263 jl
17265 jr
17266 jt
17267 jb These options set the amount of "junk" to ignore at the left,
17268 right, top, and bottom of the image, respectively. Left and right
17269 are in units of 8 pixels, while top and bottom are in units of 2
17270 lines. The default is 8 pixels on each side.
17271 sb Set the strict breaks. Setting this option to 1 will reduce the
17273 chances of filter generating an occasional mismatched frame, but it
17274 may also cause an excessive number of frames to be dropped during
17275 high motion sequences. Conversely, setting it to -1 will make
17276 filter match fields more easily. This may help processing of video
17277 where there is slight blurring between the fields, but may also
17278 cause there to be interlaced frames in the output. Default value
17279 is 0.
17280 mp Set the metric plane to use. It accepts the following values:
17282 l Use luma plane.
17284 u Use chroma blue plane.
17286 v Use chroma red plane.
17288 This option may be set to use chroma plane instead of the default
17290 luma plane for doing filter's computations. This may improve
17291 accuracy on very clean source material, but more likely will
17292 decrease accuracy, especially if there is chroma noise (rainbow
17293 effect) or any grayscale video. The main purpose of setting mp to
17294 a chroma plane is to reduce CPU load and make pullup usable in
17295 realtime on slow machines.
17296 For best results (without duplicated frames in the output file) it is
17298 necessary to change the output frame rate. For example, to inverse
17299 telecine NTSC input:
17300 ffmpeg -i input -vf pullup -r 24000/1001 ...
17302 qp
17304 Change video quantization parameters (QP).
17305 The filter accepts the following option:
17307 qp Set expression for quantization parameter.
17309 The expression is evaluated through the eval API and can contain, among
17311 others, the following constants:
17312 known
17314 1 if index is not 129, 0 otherwise.
17315 qp Sequential index starting from -129 to 128.
17317 Examples
17319 • Some equation like:
17321 qp=2+2*sin(PI*qp)
17323 random
17325 Flush video frames from internal cache of frames into a random order.
17326 No frame is discarded. Inspired by frei0r nervous filter.
17327 frames
17329 Set size in number of frames of internal cache, in range from 2 to
17330 512. Default is 30.
17331 seed
17333 Set seed for random number generator, must be an integer included
17334 between 0 and "UINT32_MAX". If not specified, or if explicitly set
17335 to less than 0, the filter will try to use a good random seed on a
17336 best effort basis.
17337 readeia608
17339 Read closed captioning (EIA-608) information from the top lines of a
17340 video frame.
17341 This filter adds frame metadata for "lavfi.readeia608.X.cc" and
17343 "lavfi.readeia608.X.line", where "X" is the number of the identified
17344 line with EIA-608 data (starting from 0). A description of each
17345 metadata value follows:
17346 lavfi.readeia608.X.cc
17348 The two bytes stored as EIA-608 data (printed in hexadecimal).
17349 lavfi.readeia608.X.line
17351 The number of the line on which the EIA-608 data was identified and
17352 read.
17353 This filter accepts the following options:
17355 scan_min
17357 Set the line to start scanning for EIA-608 data. Default is 0.
17358 scan_max
17360 Set the line to end scanning for EIA-608 data. Default is 29.
17361 spw Set the ratio of width reserved for sync code detection. Default
17363 is 0.27. Allowed range is "[0.1 - 0.7]".
17364 chp Enable checking the parity bit. In the event of a parity error, the
17366 filter will output 0x00 for that character. Default is false.
17367 lp Lowpass lines prior to further processing. Default is enabled.
17369 Commands
17371 This filter supports the all above options as commands.
17373 Examples
17375 • Output a csv with presentation time and the first two lines of
17377 identified EIA-608 captioning data.
17378 ffprobe -f lavfi -i movie=captioned_video.mov,readeia608 -show_entries frame=pts_time:frame_tags=lavfi.readeia608.0.cc,lavfi.readeia608.1.cc -of csv
17380 readvitc
17382 Read vertical interval timecode (VITC) information from the top lines
17383 of a video frame.
17384 The filter adds frame metadata key "lavfi.readvitc.tc_str" with the
17386 timecode value, if a valid timecode has been detected. Further metadata
17387 key "lavfi.readvitc.found" is set to 0/1 depending on whether timecode
17388 data has been found or not.
17389 This filter accepts the following options:
17391 scan_max
17393 Set the maximum number of lines to scan for VITC data. If the value
17394 is set to -1 the full video frame is scanned. Default is 45.
17395 thr_b
17397 Set the luma threshold for black. Accepts float numbers in the
17398 range [0.0,1.0], default value is 0.2. The value must be equal or
17399 less than "thr_w".
17400 thr_w
17402 Set the luma threshold for white. Accepts float numbers in the
17403 range [0.0,1.0], default value is 0.6. The value must be equal or
17404 greater than "thr_b".
17405 Examples
17407 • Detect and draw VITC data onto the video frame; if no valid VITC is
17409 detected, draw "--:--:--:--" as a placeholder:
17410 ffmpeg -i input.avi -filter:v 'readvitc,drawtext=fontfile=FreeMono.ttf:text=%{metadata\\:lavfi.readvitc.tc_str\\:--\\\\\\:--\\\\\\:--\\\\\\:--}:x=(w-tw)/2:y=400-ascent'
17412 remap
17414 Remap pixels using 2nd: Xmap and 3rd: Ymap input video stream.
17415 Destination pixel at position (X, Y) will be picked from source (x, y)
17417 position where x = Xmap(X, Y) and y = Ymap(X, Y). If mapping values are
17418 out of range, zero value for pixel will be used for destination pixel.
17419 Xmap and Ymap input video streams must be of same dimensions. Output
17421 video stream will have Xmap/Ymap video stream dimensions. Xmap and
17422 Ymap input video streams are 16bit depth, single channel.
17423 format
17425 Specify pixel format of output from this filter. Can be "color" or
17426 "gray". Default is "color".
17427 fill
17429 Specify the color of the unmapped pixels. For the syntax of this
17430 option, check the "Color" section in the ffmpeg-utils manual.
17431 Default color is "black".
17432 removegrain
17434 The removegrain filter is a spatial denoiser for progressive video.
17435 m0 Set mode for the first plane.
17437 m1 Set mode for the second plane.
17439 m2 Set mode for the third plane.
17441 m3 Set mode for the fourth plane.
17443 Range of mode is from 0 to 24. Description of each mode follows:
17445 0 Leave input plane unchanged. Default.
17447 1 Clips the pixel with the minimum and maximum of the 8 neighbour
17449 pixels.
17450 2 Clips the pixel with the second minimum and maximum of the 8
17452 neighbour pixels.
17453 3 Clips the pixel with the third minimum and maximum of the 8
17455 neighbour pixels.
17456 4 Clips the pixel with the fourth minimum and maximum of the 8
17458 neighbour pixels. This is equivalent to a median filter.
17459 5 Line-sensitive clipping giving the minimal change.
17461 6 Line-sensitive clipping, intermediate.
17463 7 Line-sensitive clipping, intermediate.
17465 8 Line-sensitive clipping, intermediate.
17467 9 Line-sensitive clipping on a line where the neighbours pixels are
17469 the closest.
17470 10 Replaces the target pixel with the closest neighbour.
17472 11 [1 2 1] horizontal and vertical kernel blur.
17474 12 Same as mode 11.
17476 13 Bob mode, interpolates top field from the line where the neighbours
17478 pixels are the closest.
17479 14 Bob mode, interpolates bottom field from the line where the
17481 neighbours pixels are the closest.
17482 15 Bob mode, interpolates top field. Same as 13 but with a more
17484 complicated interpolation formula.
17485 16 Bob mode, interpolates bottom field. Same as 14 but with a more
17487 complicated interpolation formula.
17488 17 Clips the pixel with the minimum and maximum of respectively the
17490 maximum and minimum of each pair of opposite neighbour pixels.
17491 18 Line-sensitive clipping using opposite neighbours whose greatest
17493 distance from the current pixel is minimal.
17494 19 Replaces the pixel with the average of its 8 neighbours.
17496 20 Averages the 9 pixels ([1 1 1] horizontal and vertical blur).
17498 21 Clips pixels using the averages of opposite neighbour.
17500 22 Same as mode 21 but simpler and faster.
17502 23 Small edge and halo removal, but reputed useless.
17504 24 Similar as 23.
17506 removelogo
17508 Suppress a TV station logo, using an image file to determine which
17509 pixels comprise the logo. It works by filling in the pixels that
17510 comprise the logo with neighboring pixels.
17511 The filter accepts the following options:
17513 filename, f
17515 Set the filter bitmap file, which can be any image format supported
17516 by libavformat. The width and height of the image file must match
17517 those of the video stream being processed.
17518 Pixels in the provided bitmap image with a value of zero are not
17520 considered part of the logo, non-zero pixels are considered part of the
17521 logo. If you use white (255) for the logo and black (0) for the rest,
17522 you will be safe. For making the filter bitmap, it is recommended to
17523 take a screen capture of a black frame with the logo visible, and then
17524 using a threshold filter followed by the erode filter once or twice.
17525 If needed, little splotches can be fixed manually. Remember that if
17527 logo pixels are not covered, the filter quality will be much reduced.
17528 Marking too many pixels as part of the logo does not hurt as much, but
17529 it will increase the amount of blurring needed to cover over the image
17530 and will destroy more information than necessary, and extra pixels will
17531 slow things down on a large logo.
17532 repeatfields
17534 This filter uses the repeat_field flag from the Video ES headers and
17535 hard repeats fields based on its value.
17536 reverse
17538 Reverse a video clip.
17539 Warning: This filter requires memory to buffer the entire clip, so
17541 trimming is suggested.
17542 Examples
17544 • Take the first 5 seconds of a clip, and reverse it.
17546 trim=end=5,reverse
17548 rgbashift
17550 Shift R/G/B/A pixels horizontally and/or vertically.
17551 The filter accepts the following options:
17553 rh Set amount to shift red horizontally.
17555 rv Set amount to shift red vertically.
17557 gh Set amount to shift green horizontally.
17559 gv Set amount to shift green vertically.
17561 bh Set amount to shift blue horizontally.
17563 bv Set amount to shift blue vertically.
17565 ah Set amount to shift alpha horizontally.
17567 av Set amount to shift alpha vertically.
17569 edge
17571 Set edge mode, can be smear, default, or warp.
17572 Commands
17574 This filter supports the all above options as commands.
17576 roberts
17578 Apply roberts cross operator to input video stream.
17579 The filter accepts the following option:
17581 planes
17583 Set which planes will be processed, unprocessed planes will be
17584 copied. By default value 0xf, all planes will be processed.
17585 scale
17587 Set value which will be multiplied with filtered result.
17588 delta
17590 Set value which will be added to filtered result.
17591 Commands
17593 This filter supports the all above options as commands.
17595 rotate
17597 Rotate video by an arbitrary angle expressed in radians.
17598 The filter accepts the following options:
17600 A description of the optional parameters follows.
17602 angle, a
17604 Set an expression for the angle by which to rotate the input video
17605 clockwise, expressed as a number of radians. A negative value will
17606 result in a counter-clockwise rotation. By default it is set to
17607 "0".
17608 This expression is evaluated for each frame.
17610 out_w, ow
17612 Set the output width expression, default value is "iw". This
17613 expression is evaluated just once during configuration.
17614 out_h, oh
17616 Set the output height expression, default value is "ih". This
17617 expression is evaluated just once during configuration.
17618 bilinear
17620 Enable bilinear interpolation if set to 1, a value of 0 disables
17621 it. Default value is 1.
17622 fillcolor, c
17624 Set the color used to fill the output area not covered by the
17625 rotated image. For the general syntax of this option, check the
17626 "Color" section in the ffmpeg-utils manual. If the special value
17627 "none" is selected then no background is printed (useful for
17628 example if the background is never shown).
17629 Default value is "black".
17631 The expressions for the angle and the output size can contain the
17633 following constants and functions:
17634 n sequential number of the input frame, starting from 0. It is always
17636 NAN before the first frame is filtered.
17637 t time in seconds of the input frame, it is set to 0 when the filter
17639 is configured. It is always NAN before the first frame is filtered.
17640 hsub
17642 vsub
17643 horizontal and vertical chroma subsample values. For example for
17644 the pixel format "yuv422p" hsub is 2 and vsub is 1.
17645 in_w, iw
17647 in_h, ih
17648 the input video width and height
17649 out_w, ow
17651 out_h, oh
17652 the output width and height, that is the size of the padded area as
17653 specified by the width and height expressions
17654 rotw(a)
17656 roth(a)
17657 the minimal width/height required for completely containing the
17658 input video rotated by a radians.
17659 These are only available when computing the out_w and out_h
17661 expressions.
17662 Examples
17664 • Rotate the input by PI/6 radians clockwise:
17666 rotate=PI/6
17668 • Rotate the input by PI/6 radians counter-clockwise:
17670 rotate=-PI/6
17672 • Rotate the input by 45 degrees clockwise:
17674 rotate=45*PI/180
17676 • Apply a constant rotation with period T, starting from an angle of
17678 PI/3:
17679 rotate=PI/3+2*PI*t/T
17681 • Make the input video rotation oscillating with a period of T
17683 seconds and an amplitude of A radians:
17684 rotate=A*sin(2*PI/T*t)
17686 • Rotate the video, output size is chosen so that the whole rotating
17688 input video is always completely contained in the output:
17689 rotate='2*PI*t:ow=hypot(iw,ih):oh=ow'
17691 • Rotate the video, reduce the output size so that no background is
17693 ever shown:
17694 rotate=2*PI*t:ow='min(iw,ih)/sqrt(2)':oh=ow:c=none
17696 Commands
17698 The filter supports the following commands:
17700 a, angle
17702 Set the angle expression. The command accepts the same syntax of
17703 the corresponding option.
17704 If the specified expression is not valid, it is kept at its current
17706 value.
17707 sab
17709 Apply Shape Adaptive Blur.
17710 The filter accepts the following options:
17712 luma_radius, lr
17714 Set luma blur filter strength, must be a value in range 0.1-4.0,
17715 default value is 1.0. A greater value will result in a more blurred
17716 image, and in slower processing.
17717 luma_pre_filter_radius, lpfr
17719 Set luma pre-filter radius, must be a value in the 0.1-2.0 range,
17720 default value is 1.0.
17721 luma_strength, ls
17723 Set luma maximum difference between pixels to still be considered,
17724 must be a value in the 0.1-100.0 range, default value is 1.0.
17725 chroma_radius, cr
17727 Set chroma blur filter strength, must be a value in range -0.9-4.0.
17728 A greater value will result in a more blurred image, and in slower
17729 processing.
17730 chroma_pre_filter_radius, cpfr
17732 Set chroma pre-filter radius, must be a value in the -0.9-2.0
17733 range.
17734 chroma_strength, cs
17736 Set chroma maximum difference between pixels to still be
17737 considered, must be a value in the -0.9-100.0 range.
17738 Each chroma option value, if not explicitly specified, is set to the
17740 corresponding luma option value.
17741 scale
17743 Scale (resize) the input video, using the libswscale library.
17744 The scale filter forces the output display aspect ratio to be the same
17746 of the input, by changing the output sample aspect ratio.
17747 If the input image format is different from the format requested by the
17749 next filter, the scale filter will convert the input to the requested
17750 format.
17751 Options
17753 The filter accepts the following options, or any of the options
17755 supported by the libswscale scaler.
17756 See the ffmpeg-scaler manual for the complete list of scaler options.
17758 width, w
17760 height, h
17761 Set the output video dimension expression. Default value is the
17762 input dimension.
17763 If the width or w value is 0, the input width is used for the
17765 output. If the height or h value is 0, the input height is used for
17766 the output.
17767 If one and only one of the values is -n with n >= 1, the scale
17769 filter will use a value that maintains the aspect ratio of the
17770 input image, calculated from the other specified dimension. After
17771 that it will, however, make sure that the calculated dimension is
17772 divisible by n and adjust the value if necessary.
17773 If both values are -n with n >= 1, the behavior will be identical
17775 to both values being set to 0 as previously detailed.
17776 See below for the list of accepted constants for use in the
17778 dimension expression.
17779 eval
17781 Specify when to evaluate width and height expression. It accepts
17782 the following values:
17783 init
17785 Only evaluate expressions once during the filter initialization
17786 or when a command is processed.
17787 frame
17789 Evaluate expressions for each incoming frame.
17790 Default value is init.
17792 interl
17794 Set the interlacing mode. It accepts the following values:
17795 1 Force interlaced aware scaling.
17797 0 Do not apply interlaced scaling.
17799 -1 Select interlaced aware scaling depending on whether the source
17801 frames are flagged as interlaced or not.
17802 Default value is 0.
17804 flags
17806 Set libswscale scaling flags. See the ffmpeg-scaler manual for the
17807 complete list of values. If not explicitly specified the filter
17808 applies the default flags.
17809 param0, param1
17811 Set libswscale input parameters for scaling algorithms that need
17812 them. See the ffmpeg-scaler manual for the complete documentation.
17813 If not explicitly specified the filter applies empty parameters.
17814 size, s
17816 Set the video size. For the syntax of this option, check the "Video
17817 size" section in the ffmpeg-utils manual.
17818 in_color_matrix
17820 out_color_matrix
17821 Set in/output YCbCr color space type.
17822 This allows the autodetected value to be overridden as well as
17824 allows forcing a specific value used for the output and encoder.
17825 If not specified, the color space type depends on the pixel format.
17827 Possible values:
17829 auto
17831 Choose automatically.
17832 bt709
17834 Format conforming to International Telecommunication Union
17835 (ITU) Recommendation BT.709.
17836 fcc Set color space conforming to the United States Federal
17838 Communications Commission (FCC) Code of Federal Regulations
17839 (CFR) Title 47 (2003) 73.682 (a).
17840 bt601
17842 bt470
17843 smpte170m
17844 Set color space conforming to:
17845 • ITU Radiocommunication Sector (ITU-R) Recommendation BT.601
17847 • ITU-R Rec. BT.470-6 (1998) Systems B, B1, and G
17849 • Society of Motion Picture and Television Engineers (SMPTE)
17851 ST 170:2004
17852 smpte240m
17854 Set color space conforming to SMPTE ST 240:1999.
17855 bt2020
17857 Set color space conforming to ITU-R BT.2020 non-constant
17858 luminance system.
17859 in_range
17861 out_range
17862 Set in/output YCbCr sample range.
17863 This allows the autodetected value to be overridden as well as
17865 allows forcing a specific value used for the output and encoder. If
17866 not specified, the range depends on the pixel format. Possible
17867 values:
17868 auto/unknown
17870 Choose automatically.
17871 jpeg/full/pc
17873 Set full range (0-255 in case of 8-bit luma).
17874 mpeg/limited/tv
17876 Set "MPEG" range (16-235 in case of 8-bit luma).
17877 force_original_aspect_ratio
17879 Enable decreasing or increasing output video width or height if
17880 necessary to keep the original aspect ratio. Possible values:
17881 disable
17883 Scale the video as specified and disable this feature.
17884 decrease
17886 The output video dimensions will automatically be decreased if
17887 needed.
17888 increase
17890 The output video dimensions will automatically be increased if
17891 needed.
17892 One useful instance of this option is that when you know a specific
17894 device's maximum allowed resolution, you can use this to limit the
17895 output video to that, while retaining the aspect ratio. For
17896 example, device A allows 1280x720 playback, and your video is
17897 1920x800. Using this option (set it to decrease) and specifying
17898 1280x720 to the command line makes the output 1280x533.
17899 Please note that this is a different thing than specifying -1 for w
17901 or h, you still need to specify the output resolution for this
17902 option to work.
17903 force_divisible_by
17905 Ensures that both the output dimensions, width and height, are
17906 divisible by the given integer when used together with
17907 force_original_aspect_ratio. This works similar to using "-n" in
17908 the w and h options.
17909 This option respects the value set for force_original_aspect_ratio,
17911 increasing or decreasing the resolution accordingly. The video's
17912 aspect ratio may be slightly modified.
17913 This option can be handy if you need to have a video fit within or
17915 exceed a defined resolution using force_original_aspect_ratio but
17916 also have encoder restrictions on width or height divisibility.
17917 The values of the w and h options are expressions containing the
17919 following constants:
17920 in_w
17922 in_h
17923 The input width and height
17924 iw
17926 ih These are the same as in_w and in_h.
17927 out_w
17929 out_h
17930 The output (scaled) width and height
17931 ow
17933 oh These are the same as out_w and out_h
17934 a The same as iw / ih
17936 sar input sample aspect ratio
17938 dar The input display aspect ratio. Calculated from "(iw / ih) * sar".
17940 hsub
17942 vsub
17943 horizontal and vertical input chroma subsample values. For example
17944 for the pixel format "yuv422p" hsub is 2 and vsub is 1.
17945 ohsub
17947 ovsub
17948 horizontal and vertical output chroma subsample values. For example
17949 for the pixel format "yuv422p" hsub is 2 and vsub is 1.
17950 n The (sequential) number of the input frame, starting from 0. Only
17952 available with "eval=frame".
17953 t The presentation timestamp of the input frame, expressed as a
17955 number of seconds. Only available with "eval=frame".
17956 pos The position (byte offset) of the frame in the input stream, or NaN
17958 if this information is unavailable and/or meaningless (for example
17959 in case of synthetic video). Only available with "eval=frame".
17960 Examples
17962 • Scale the input video to a size of 200x100
17964 scale=w=200:h=100
17966 This is equivalent to:
17968 scale=200:100
17970 or:
17972 scale=200x100
17974 • Specify a size abbreviation for the output size:
17976 scale=qcif
17978 which can also be written as:
17980 scale=size=qcif
17982 • Scale the input to 2x:
17984 scale=w=2*iw:h=2*ih
17986 • The above is the same as:
17988 scale=2*in_w:2*in_h
17990 • Scale the input to 2x with forced interlaced scaling:
17992 scale=2*iw:2*ih:interl=1
17994 • Scale the input to half size:
17996 scale=w=iw/2:h=ih/2
17998 • Increase the width, and set the height to the same size:
18000 scale=3/2*iw:ow
18002 • Seek Greek harmony:
18004 scale=iw:1/PHI*iw
18006 scale=ih*PHI:ih
18007 • Increase the height, and set the width to 3/2 of the height:
18009 scale=w=3/2*oh:h=3/5*ih
18011 • Increase the size, making the size a multiple of the chroma
18013 subsample values:
18014 scale="trunc(3/2*iw/hsub)*hsub:trunc(3/2*ih/vsub)*vsub"
18016 • Increase the width to a maximum of 500 pixels, keeping the same
18018 aspect ratio as the input:
18019 scale=w='min(500\, iw*3/2):h=-1'
18021 • Make pixels square by combining scale and setsar:
18023 scale='trunc(ih*dar):ih',setsar=1/1
18025 • Make pixels square by combining scale and setsar, making sure the
18027 resulting resolution is even (required by some codecs):
18028 scale='trunc(ih*dar/2)*2:trunc(ih/2)*2',setsar=1/1
18030 Commands
18032 This filter supports the following commands:
18034 width, w
18036 height, h
18037 Set the output video dimension expression. The command accepts the
18038 same syntax of the corresponding option.
18039 If the specified expression is not valid, it is kept at its current
18041 value.
18042 scale_cuda
18044 Scale (resize) and convert (pixel format) the input video, using
18045 accelerated CUDA kernels. Setting the output width and height works in
18046 the same way as for the scale filter.
18047 The filter accepts the following options:
18049 w
18051 h Set the output video dimension expression. Default value is the
18052 input dimension.
18053 Allows for the same expressions as the scale filter.
18055 interp_algo
18057 Sets the algorithm used for scaling:
18058 nearest
18060 Nearest neighbour
18061 Used by default if input parameters match the desired output.
18063 bilinear
18065 Bilinear
18066 bicubic
18068 Bicubic
18069 This is the default.
18071 lanczos
18073 Lanczos
18074 format
18076 Controls the output pixel format. By default, or if none is
18077 specified, the input pixel format is used.
18078 The filter does not support converting between YUV and RGB pixel
18080 formats.
18081 passthrough
18083 If set to 0, every frame is processed, even if no conversion is
18084 neccesary. This mode can be useful to use the filter as a buffer
18085 for a downstream frame-consumer that exhausts the limited decoder
18086 frame pool.
18087 If set to 1, frames are passed through as-is if they match the
18089 desired output parameters. This is the default behaviour.
18090 param
18092 Algorithm-Specific parameter.
18093 Affects the curves of the bicubic algorithm.
18095 force_original_aspect_ratio
18097 force_divisible_by
18098 Work the same as the identical scale filter options.
18099 Examples
18101 • Scale input to 720p, keeping aspect ratio and ensuring the output
18103 is yuv420p.
18104 scale_cuda=-2:720:format=yuv420p
18106 • Upscale to 4K using nearest neighbour algorithm.
18108 scale_cuda=4096:2160:interp_algo=nearest
18110 • Don't do any conversion or scaling, but copy all input frames into
18112 newly allocated ones. This can be useful to deal with a filter and
18113 encode chain that otherwise exhausts the decoders frame pool.
18114 scale_cuda=passthrough=0
18116 scale_npp
18118 Use the NVIDIA Performance Primitives (libnpp) to perform scaling
18119 and/or pixel format conversion on CUDA video frames. Setting the output
18120 width and height works in the same way as for the scale filter.
18121 The following additional options are accepted:
18123 format
18125 The pixel format of the output CUDA frames. If set to the string
18126 "same" (the default), the input format will be kept. Note that
18127 automatic format negotiation and conversion is not yet supported
18128 for hardware frames
18129 interp_algo
18131 The interpolation algorithm used for resizing. One of the
18132 following:
18133 nn Nearest neighbour.
18135 linear
18137 cubic
18138 cubic2p_bspline
18139 2-parameter cubic (B=1, C=0)
18140 cubic2p_catmullrom
18142 2-parameter cubic (B=0, C=1/2)
18143 cubic2p_b05c03
18145 2-parameter cubic (B=1/2, C=3/10)
18146 super
18148 Supersampling
18149 lanczos
18151 force_original_aspect_ratio
18152 Enable decreasing or increasing output video width or height if
18153 necessary to keep the original aspect ratio. Possible values:
18154 disable
18156 Scale the video as specified and disable this feature.
18157 decrease
18159 The output video dimensions will automatically be decreased if
18160 needed.
18161 increase
18163 The output video dimensions will automatically be increased if
18164 needed.
18165 One useful instance of this option is that when you know a specific
18167 device's maximum allowed resolution, you can use this to limit the
18168 output video to that, while retaining the aspect ratio. For
18169 example, device A allows 1280x720 playback, and your video is
18170 1920x800. Using this option (set it to decrease) and specifying
18171 1280x720 to the command line makes the output 1280x533.
18172 Please note that this is a different thing than specifying -1 for w
18174 or h, you still need to specify the output resolution for this
18175 option to work.
18176 force_divisible_by
18178 Ensures that both the output dimensions, width and height, are
18179 divisible by the given integer when used together with
18180 force_original_aspect_ratio. This works similar to using "-n" in
18181 the w and h options.
18182 This option respects the value set for force_original_aspect_ratio,
18184 increasing or decreasing the resolution accordingly. The video's
18185 aspect ratio may be slightly modified.
18186 This option can be handy if you need to have a video fit within or
18188 exceed a defined resolution using force_original_aspect_ratio but
18189 also have encoder restrictions on width or height divisibility.
18190 eval
18192 Specify when to evaluate width and height expression. It accepts
18193 the following values:
18194 init
18196 Only evaluate expressions once during the filter initialization
18197 or when a command is processed.
18198 frame
18200 Evaluate expressions for each incoming frame.
18201 The values of the w and h options are expressions containing the
18203 following constants:
18204 in_w
18206 in_h
18207 The input width and height
18208 iw
18210 ih These are the same as in_w and in_h.
18211 out_w
18213 out_h
18214 The output (scaled) width and height
18215 ow
18217 oh These are the same as out_w and out_h
18218 a The same as iw / ih
18220 sar input sample aspect ratio
18222 dar The input display aspect ratio. Calculated from "(iw / ih) * sar".
18224 n The (sequential) number of the input frame, starting from 0. Only
18226 available with "eval=frame".
18227 t The presentation timestamp of the input frame, expressed as a
18229 number of seconds. Only available with "eval=frame".
18230 pos The position (byte offset) of the frame in the input stream, or NaN
18232 if this information is unavailable and/or meaningless (for example
18233 in case of synthetic video). Only available with "eval=frame".
18234 scale2ref
18236 Scale (resize) the input video, based on a reference video.
18237 See the scale filter for available options, scale2ref supports the same
18239 but uses the reference video instead of the main input as basis.
18240 scale2ref also supports the following additional constants for the w
18241 and h options:
18242 main_w
18244 main_h
18245 The main input video's width and height
18246 main_a
18248 The same as main_w / main_h
18249 main_sar
18251 The main input video's sample aspect ratio
18252 main_dar, mdar
18254 The main input video's display aspect ratio. Calculated from
18255 "(main_w / main_h) * main_sar".
18256 main_hsub
18258 main_vsub
18259 The main input video's horizontal and vertical chroma subsample
18260 values. For example for the pixel format "yuv422p" hsub is 2 and
18261 vsub is 1.
18262 main_n
18264 The (sequential) number of the main input frame, starting from 0.
18265 Only available with "eval=frame".
18266 main_t
18268 The presentation timestamp of the main input frame, expressed as a
18269 number of seconds. Only available with "eval=frame".
18270 main_pos
18272 The position (byte offset) of the frame in the main input stream,
18273 or NaN if this information is unavailable and/or meaningless (for
18274 example in case of synthetic video). Only available with
18275 "eval=frame".
18276 Examples
18278 • Scale a subtitle stream (b) to match the main video (a) in size
18280 before overlaying
18281 'scale2ref[b][a];[a][b]overlay'
18283 • Scale a logo to 1/10th the height of a video, while preserving its
18285 display aspect ratio.
18286 [logo-in][video-in]scale2ref=w=oh*mdar:h=ih/10[logo-out][video-out]
18288 Commands
18290 This filter supports the following commands:
18292 width, w
18294 height, h
18295 Set the output video dimension expression. The command accepts the
18296 same syntax of the corresponding option.
18297 If the specified expression is not valid, it is kept at its current
18299 value.
18300 scale2ref_npp
18302 Use the NVIDIA Performance Primitives (libnpp) to scale (resize) the
18303 input video, based on a reference video.
18304 See the scale_npp filter for available options, scale2ref_npp supports
18306 the same but uses the reference video instead of the main input as
18307 basis. scale2ref_npp also supports the following additional constants
18308 for the w and h options:
18309 main_w
18311 main_h
18312 The main input video's width and height
18313 main_a
18315 The same as main_w / main_h
18316 main_sar
18318 The main input video's sample aspect ratio
18319 main_dar, mdar
18321 The main input video's display aspect ratio. Calculated from
18322 "(main_w / main_h) * main_sar".
18323 main_n
18325 The (sequential) number of the main input frame, starting from 0.
18326 Only available with "eval=frame".
18327 main_t
18329 The presentation timestamp of the main input frame, expressed as a
18330 number of seconds. Only available with "eval=frame".
18331 main_pos
18333 The position (byte offset) of the frame in the main input stream,
18334 or NaN if this information is unavailable and/or meaningless (for
18335 example in case of synthetic video). Only available with
18336 "eval=frame".
18337 Examples
18339 • Scale a subtitle stream (b) to match the main video (a) in size
18341 before overlaying
18342 'scale2ref_npp[b][a];[a][b]overlay_cuda'
18344 • Scale a logo to 1/10th the height of a video, while preserving its
18346 display aspect ratio.
18347 [logo-in][video-in]scale2ref_npp=w=oh*mdar:h=ih/10[logo-out][video-out]
18349 scharr
18351 Apply scharr operator to input video stream.
18352 The filter accepts the following option:
18354 planes
18356 Set which planes will be processed, unprocessed planes will be
18357 copied. By default value 0xf, all planes will be processed.
18358 scale
18360 Set value which will be multiplied with filtered result.
18361 delta
18363 Set value which will be added to filtered result.
18364 Commands
18366 This filter supports the all above options as commands.
18368 scroll
18370 Scroll input video horizontally and/or vertically by constant speed.
18371 The filter accepts the following options:
18373 horizontal, h
18375 Set the horizontal scrolling speed. Default is 0. Allowed range is
18376 from -1 to 1. Negative values changes scrolling direction.
18377 vertical, v
18379 Set the vertical scrolling speed. Default is 0. Allowed range is
18380 from -1 to 1. Negative values changes scrolling direction.
18381 hpos
18383 Set the initial horizontal scrolling position. Default is 0.
18384 Allowed range is from 0 to 1.
18385 vpos
18387 Set the initial vertical scrolling position. Default is 0. Allowed
18388 range is from 0 to 1.
18389 Commands
18391 This filter supports the following commands:
18393 horizontal, h
18395 Set the horizontal scrolling speed.
18396 vertical, v
18398 Set the vertical scrolling speed.
18399 scdet
18401 Detect video scene change.
18402 This filter sets frame metadata with mafd between frame, the scene
18404 score, and forward the frame to the next filter, so they can use these
18405 metadata to detect scene change or others.
18406 In addition, this filter logs a message and sets frame metadata when it
18408 detects a scene change by threshold.
18409 "lavfi.scd.mafd" metadata keys are set with mafd for every frame.
18411 "lavfi.scd.score" metadata keys are set with scene change score for
18413 every frame to detect scene change.
18414 "lavfi.scd.time" metadata keys are set with current filtered frame time
18416 which detect scene change with threshold.
18417 The filter accepts the following options:
18419 threshold, t
18421 Set the scene change detection threshold as a percentage of maximum
18422 change. Good values are in the "[8.0, 14.0]" range. The range for
18423 threshold is "[0., 100.]".
18424 Default value is 10..
18426 sc_pass, s
18428 Set the flag to pass scene change frames to the next filter.
18429 Default value is 0 You can enable it if you want to get snapshot of
18430 scene change frames only.
18431 selectivecolor
18433 Adjust cyan, magenta, yellow and black (CMYK) to certain ranges of
18434 colors (such as "reds", "yellows", "greens", "cyans", ...). The
18435 adjustment range is defined by the "purity" of the color (that is, how
18436 saturated it already is).
18437 This filter is similar to the Adobe Photoshop Selective Color tool.
18439 The filter accepts the following options:
18441 correction_method
18443 Select color correction method.
18444 Available values are:
18446 absolute
18448 Specified adjustments are applied "as-is" (added/subtracted to
18449 original pixel component value).
18450 relative
18452 Specified adjustments are relative to the original component
18453 value.
18454 Default is "absolute".
18456 reds
18458 Adjustments for red pixels (pixels where the red component is the
18459 maximum)
18460 yellows
18462 Adjustments for yellow pixels (pixels where the blue component is
18463 the minimum)
18464 greens
18466 Adjustments for green pixels (pixels where the green component is
18467 the maximum)
18468 cyans
18470 Adjustments for cyan pixels (pixels where the red component is the
18471 minimum)
18472 blues
18474 Adjustments for blue pixels (pixels where the blue component is the
18475 maximum)
18476 magentas
18478 Adjustments for magenta pixels (pixels where the green component is
18479 the minimum)
18480 whites
18482 Adjustments for white pixels (pixels where all components are
18483 greater than 128)
18484 neutrals
18486 Adjustments for all pixels except pure black and pure white
18487 blacks
18489 Adjustments for black pixels (pixels where all components are
18490 lesser than 128)
18491 psfile
18493 Specify a Photoshop selective color file (".asv") to import the
18494 settings from.
18495 All the adjustment settings (reds, yellows, ...) accept up to 4 space
18497 separated floating point adjustment values in the [-1,1] range,
18498 respectively to adjust the amount of cyan, magenta, yellow and black
18499 for the pixels of its range.
18500 Examples
18502 • Increase cyan by 50% and reduce yellow by 33% in every green areas,
18504 and increase magenta by 27% in blue areas:
18505 selectivecolor=greens=.5 0 -.33 0:blues=0 .27
18507 • Use a Photoshop selective color preset:
18509 selectivecolor=psfile=MySelectiveColorPresets/Misty.asv
18511 separatefields
18513 The "separatefields" takes a frame-based video input and splits each
18514 frame into its components fields, producing a new half height clip with
18515 twice the frame rate and twice the frame count.
18516 This filter use field-dominance information in frame to decide which of
18518 each pair of fields to place first in the output. If it gets it wrong
18519 use setfield filter before "separatefields" filter.
18520 setdar, setsar
18522 The "setdar" filter sets the Display Aspect Ratio for the filter output
18523 video.
18524 This is done by changing the specified Sample (aka Pixel) Aspect Ratio,
18526 according to the following equation:
18527 <DAR> = <HORIZONTAL_RESOLUTION> / <VERTICAL_RESOLUTION> * <SAR>
18529 Keep in mind that the "setdar" filter does not modify the pixel
18531 dimensions of the video frame. Also, the display aspect ratio set by
18532 this filter may be changed by later filters in the filterchain, e.g. in
18533 case of scaling or if another "setdar" or a "setsar" filter is applied.
18534 The "setsar" filter sets the Sample (aka Pixel) Aspect Ratio for the
18536 filter output video.
18537 Note that as a consequence of the application of this filter, the
18539 output display aspect ratio will change according to the equation
18540 above.
18541 Keep in mind that the sample aspect ratio set by the "setsar" filter
18543 may be changed by later filters in the filterchain, e.g. if another
18544 "setsar" or a "setdar" filter is applied.
18545 It accepts the following parameters:
18547 r, ratio, dar ("setdar" only), sar ("setsar" only)
18549 Set the aspect ratio used by the filter.
18550 The parameter can be a floating point number string, an expression,
18552 or a string of the form num:den, where num and den are the
18553 numerator and denominator of the aspect ratio. If the parameter is
18554 not specified, it is assumed the value "0". In case the form
18555 "num:den" is used, the ":" character should be escaped.
18556 max Set the maximum integer value to use for expressing numerator and
18558 denominator when reducing the expressed aspect ratio to a rational.
18559 Default value is 100.
18560 The parameter sar is an expression containing the following constants:
18562 E, PI, PHI
18564 These are approximated values for the mathematical constants e
18565 (Euler's number), pi (Greek pi), and phi (the golden ratio).
18566 w, h
18568 The input width and height.
18569 a These are the same as w / h.
18571 sar The input sample aspect ratio.
18573 dar The input display aspect ratio. It is the same as (w / h) * sar.
18575 hsub, vsub
18577 Horizontal and vertical chroma subsample values. For example, for
18578 the pixel format "yuv422p" hsub is 2 and vsub is 1.
18579 Examples
18581 • To change the display aspect ratio to 16:9, specify one of the
18583 following:
18584 setdar=dar=1.77777
18586 setdar=dar=16/9
18587 • To change the sample aspect ratio to 10:11, specify:
18589 setsar=sar=10/11
18591 • To set a display aspect ratio of 16:9, and specify a maximum
18593 integer value of 1000 in the aspect ratio reduction, use the
18594 command:
18595 setdar=ratio=16/9:max=1000
18597 setfield
18599 Force field for the output video frame.
18600 The "setfield" filter marks the interlace type field for the output
18602 frames. It does not change the input frame, but only sets the
18603 corresponding property, which affects how the frame is treated by
18604 following filters (e.g. "fieldorder" or "yadif").
18605 The filter accepts the following options:
18607 mode
18609 Available values are:
18610 auto
18612 Keep the same field property.
18613 bff Mark the frame as bottom-field-first.
18615 tff Mark the frame as top-field-first.
18617 prog
18619 Mark the frame as progressive.
18620 setparams
18622 Force frame parameter for the output video frame.
18623 The "setparams" filter marks interlace and color range for the output
18625 frames. It does not change the input frame, but only sets the
18626 corresponding property, which affects how the frame is treated by
18627 filters/encoders.
18628 field_mode
18630 Available values are:
18631 auto
18633 Keep the same field property (default).
18634 bff Mark the frame as bottom-field-first.
18636 tff Mark the frame as top-field-first.
18638 prog
18640 Mark the frame as progressive.
18641 range
18643 Available values are:
18644 auto
18646 Keep the same color range property (default).
18647 unspecified, unknown
18649 Mark the frame as unspecified color range.
18650 limited, tv, mpeg
18652 Mark the frame as limited range.
18653 full, pc, jpeg
18655 Mark the frame as full range.
18656 color_primaries
18658 Set the color primaries. Available values are:
18659 auto
18661 Keep the same color primaries property (default).
18662 bt709
18664 unknown
18665 bt470m
18666 bt470bg
18667 smpte170m
18668 smpte240m
18669 film
18670 bt2020
18671 smpte428
18672 smpte431
18673 smpte432
18674 jedec-p22
18675 color_trc
18676 Set the color transfer. Available values are:
18677 auto
18679 Keep the same color trc property (default).
18680 bt709
18682 unknown
18683 bt470m
18684 bt470bg
18685 smpte170m
18686 smpte240m
18687 linear
18688 log100
18689 log316
18690 iec61966-2-4
18691 bt1361e
18692 iec61966-2-1
18693 bt2020-10
18694 bt2020-12
18695 smpte2084
18696 smpte428
18697 arib-std-b67
18698 colorspace
18699 Set the colorspace. Available values are:
18700 auto
18702 Keep the same colorspace property (default).
18703 gbr
18705 bt709
18706 unknown
18707 fcc
18708 bt470bg
18709 smpte170m
18710 smpte240m
18711 ycgco
18712 bt2020nc
18713 bt2020c
18714 smpte2085
18715 chroma-derived-nc
18716 chroma-derived-c
18717 ictcp
18718 sharpen_npp
18720 Use the NVIDIA Performance Primitives (libnpp) to perform image
18721 sharpening with border control.
18722 The following additional options are accepted:
18724 border_type
18726 Type of sampling to be used ad frame borders. One of the following:
18727 replicate
18729 Replicate pixel values.
18730 shear
18732 Apply shear transform to input video.
18733 This filter supports the following options:
18735 shx Shear factor in X-direction. Default value is 0. Allowed range is
18737 from -2 to 2.
18738 shy Shear factor in Y-direction. Default value is 0. Allowed range is
18740 from -2 to 2.
18741 fillcolor, c
18743 Set the color used to fill the output area not covered by the
18744 transformed video. For the general syntax of this option, check the
18745 "Color" section in the ffmpeg-utils manual. If the special value
18746 "none" is selected then no background is printed (useful for
18747 example if the background is never shown).
18748 Default value is "black".
18750 interp
18752 Set interpolation type. Can be "bilinear" or "nearest". Default is
18753 "bilinear".
18754 Commands
18756 This filter supports the all above options as commands.
18758 showinfo
18760 Show a line containing various information for each input video frame.
18761 The input video is not modified.
18762 This filter supports the following options:
18764 checksum
18766 Calculate checksums of each plane. By default enabled.
18767 The shown line contains a sequence of key/value pairs of the form
18769 key:value.
18770 The following values are shown in the output:
18772 n The (sequential) number of the input frame, starting from 0.
18774 pts The Presentation TimeStamp of the input frame, expressed as a
18776 number of time base units. The time base unit depends on the filter
18777 input pad.
18778 pts_time
18780 The Presentation TimeStamp of the input frame, expressed as a
18781 number of seconds.
18782 pos The position of the frame in the input stream, or -1 if this
18784 information is unavailable and/or meaningless (for example in case
18785 of synthetic video).
18786 fmt The pixel format name.
18788 sar The sample aspect ratio of the input frame, expressed in the form
18790 num/den.
18791 s The size of the input frame. For the syntax of this option, check
18793 the "Video size" section in the ffmpeg-utils manual.
18794 i The type of interlaced mode ("P" for "progressive", "T" for top
18796 field first, "B" for bottom field first).
18797 iskey
18799 This is 1 if the frame is a key frame, 0 otherwise.
18800 type
18802 The picture type of the input frame ("I" for an I-frame, "P" for a
18803 P-frame, "B" for a B-frame, or "?" for an unknown type). Also
18804 refer to the documentation of the "AVPictureType" enum and of the
18805 "av_get_picture_type_char" function defined in libavutil/avutil.h.
18806 checksum
18808 The Adler-32 checksum (printed in hexadecimal) of all the planes of
18809 the input frame.
18810 plane_checksum
18812 The Adler-32 checksum (printed in hexadecimal) of each plane of the
18813 input frame, expressed in the form "[c0 c1 c2 c3]".
18814 mean
18816 The mean value of pixels in each plane of the input frame,
18817 expressed in the form "[mean0 mean1 mean2 mean3]".
18818 stdev
18820 The standard deviation of pixel values in each plane of the input
18821 frame, expressed in the form "[stdev0 stdev1 stdev2 stdev3]".
18822 showpalette
18824 Displays the 256 colors palette of each frame. This filter is only
18825 relevant for pal8 pixel format frames.
18826 It accepts the following option:
18828 s Set the size of the box used to represent one palette color entry.
18830 Default is 30 (for a "30x30" pixel box).
18831 shuffleframes
18833 Reorder and/or duplicate and/or drop video frames.
18834 It accepts the following parameters:
18836 mapping
18838 Set the destination indexes of input frames. This is space or '|'
18839 separated list of indexes that maps input frames to output frames.
18840 Number of indexes also sets maximal value that each index may have.
18841 '-1' index have special meaning and that is to drop frame.
18842 The first frame has the index 0. The default is to keep the input
18844 unchanged.
18845 Examples
18847 • Swap second and third frame of every three frames of the input:
18849 ffmpeg -i INPUT -vf "shuffleframes=0 2 1" OUTPUT
18851 • Swap 10th and 1st frame of every ten frames of the input:
18853 ffmpeg -i INPUT -vf "shuffleframes=9 1 2 3 4 5 6 7 8 0" OUTPUT
18855 shufflepixels
18857 Reorder pixels in video frames.
18858 This filter accepts the following options:
18860 direction, d
18862 Set shuffle direction. Can be forward or inverse direction.
18863 Default direction is forward.
18864 mode, m
18866 Set shuffle mode. Can be horizontal, vertical or block mode.
18867 width, w
18869 height, h
18870 Set shuffle block_size. In case of horizontal shuffle mode only
18871 width part of size is used, and in case of vertical shuffle mode
18872 only height part of size is used.
18873 seed, s
18875 Set random seed used with shuffling pixels. Mainly useful to set to
18876 be able to reverse filtering process to get original input. For
18877 example, to reverse forward shuffle you need to use same parameters
18878 and exact same seed and to set direction to inverse.
18879 shuffleplanes
18881 Reorder and/or duplicate video planes.
18882 It accepts the following parameters:
18884 map0
18886 The index of the input plane to be used as the first output plane.
18887 map1
18889 The index of the input plane to be used as the second output plane.
18890 map2
18892 The index of the input plane to be used as the third output plane.
18893 map3
18895 The index of the input plane to be used as the fourth output plane.
18896 The first plane has the index 0. The default is to keep the input
18898 unchanged.
18899 Examples
18901 • Swap the second and third planes of the input:
18903 ffmpeg -i INPUT -vf shuffleplanes=0:2:1:3 OUTPUT
18905 signalstats
18907 Evaluate various visual metrics that assist in determining issues
18908 associated with the digitization of analog video media.
18909 By default the filter will log these metadata values:
18911 YMIN
18913 Display the minimal Y value contained within the input frame.
18914 Expressed in range of [0-255].
18915 YLOW
18917 Display the Y value at the 10% percentile within the input frame.
18918 Expressed in range of [0-255].
18919 YAVG
18921 Display the average Y value within the input frame. Expressed in
18922 range of [0-255].
18923 YHIGH
18925 Display the Y value at the 90% percentile within the input frame.
18926 Expressed in range of [0-255].
18927 YMAX
18929 Display the maximum Y value contained within the input frame.
18930 Expressed in range of [0-255].
18931 UMIN
18933 Display the minimal U value contained within the input frame.
18934 Expressed in range of [0-255].
18935 ULOW
18937 Display the U value at the 10% percentile within the input frame.
18938 Expressed in range of [0-255].
18939 UAVG
18941 Display the average U value within the input frame. Expressed in
18942 range of [0-255].
18943 UHIGH
18945 Display the U value at the 90% percentile within the input frame.
18946 Expressed in range of [0-255].
18947 UMAX
18949 Display the maximum U value contained within the input frame.
18950 Expressed in range of [0-255].
18951 VMIN
18953 Display the minimal V value contained within the input frame.
18954 Expressed in range of [0-255].
18955 VLOW
18957 Display the V value at the 10% percentile within the input frame.
18958 Expressed in range of [0-255].
18959 VAVG
18961 Display the average V value within the input frame. Expressed in
18962 range of [0-255].
18963 VHIGH
18965 Display the V value at the 90% percentile within the input frame.
18966 Expressed in range of [0-255].
18967 VMAX
18969 Display the maximum V value contained within the input frame.
18970 Expressed in range of [0-255].
18971 SATMIN
18973 Display the minimal saturation value contained within the input
18974 frame. Expressed in range of [0-~181.02].
18975 SATLOW
18977 Display the saturation value at the 10% percentile within the input
18978 frame. Expressed in range of [0-~181.02].
18979 SATAVG
18981 Display the average saturation value within the input frame.
18982 Expressed in range of [0-~181.02].
18983 SATHIGH
18985 Display the saturation value at the 90% percentile within the input
18986 frame. Expressed in range of [0-~181.02].
18987 SATMAX
18989 Display the maximum saturation value contained within the input
18990 frame. Expressed in range of [0-~181.02].
18991 HUEMED
18993 Display the median value for hue within the input frame. Expressed
18994 in range of [0-360].
18995 HUEAVG
18997 Display the average value for hue within the input frame. Expressed
18998 in range of [0-360].
18999 YDIF
19001 Display the average of sample value difference between all values
19002 of the Y plane in the current frame and corresponding values of the
19003 previous input frame. Expressed in range of [0-255].
19004 UDIF
19006 Display the average of sample value difference between all values
19007 of the U plane in the current frame and corresponding values of the
19008 previous input frame. Expressed in range of [0-255].
19009 VDIF
19011 Display the average of sample value difference between all values
19012 of the V plane in the current frame and corresponding values of the
19013 previous input frame. Expressed in range of [0-255].
19014 YBITDEPTH
19016 Display bit depth of Y plane in current frame. Expressed in range
19017 of [0-16].
19018 UBITDEPTH
19020 Display bit depth of U plane in current frame. Expressed in range
19021 of [0-16].
19022 VBITDEPTH
19024 Display bit depth of V plane in current frame. Expressed in range
19025 of [0-16].
19026 The filter accepts the following options:
19028 stat
19030 out stat specify an additional form of image analysis. out output
19031 video with the specified type of pixel highlighted.
19032 Both options accept the following values:
19034 tout
19036 Identify temporal outliers pixels. A temporal outlier is a
19037 pixel unlike the neighboring pixels of the same field. Examples
19038 of temporal outliers include the results of video dropouts,
19039 head clogs, or tape tracking issues.
19040 vrep
19042 Identify vertical line repetition. Vertical line repetition
19043 includes similar rows of pixels within a frame. In born-digital
19044 video vertical line repetition is common, but this pattern is
19045 uncommon in video digitized from an analog source. When it
19046 occurs in video that results from the digitization of an analog
19047 source it can indicate concealment from a dropout compensator.
19048 brng
19050 Identify pixels that fall outside of legal broadcast range.
19051 color, c
19053 Set the highlight color for the out option. The default color is
19054 yellow.
19055 Examples
19057 • Output data of various video metrics:
19059 ffprobe -f lavfi movie=example.mov,signalstats="stat=tout+vrep+brng" -show_frames
19061 • Output specific data about the minimum and maximum values of the Y
19063 plane per frame:
19064 ffprobe -f lavfi movie=example.mov,signalstats -show_entries frame_tags=lavfi.signalstats.YMAX,lavfi.signalstats.YMIN
19066 • Playback video while highlighting pixels that are outside of
19068 broadcast range in red.
19069 ffplay example.mov -vf signalstats="out=brng:color=red"
19071 • Playback video with signalstats metadata drawn over the frame.
19073 ffplay example.mov -vf signalstats=stat=brng+vrep+tout,drawtext=fontfile=FreeSerif.ttf:textfile=signalstat_drawtext.txt
19075 The contents of signalstat_drawtext.txt used in the command are:
19077 time %{pts:hms}
19079 Y (%{metadata:lavfi.signalstats.YMIN}-%{metadata:lavfi.signalstats.YMAX})
19080 U (%{metadata:lavfi.signalstats.UMIN}-%{metadata:lavfi.signalstats.UMAX})
19081 V (%{metadata:lavfi.signalstats.VMIN}-%{metadata:lavfi.signalstats.VMAX})
19082 saturation maximum: %{metadata:lavfi.signalstats.SATMAX}
19083 signature
19085 Calculates the MPEG-7 Video Signature. The filter can handle more than
19086 one input. In this case the matching between the inputs can be
19087 calculated additionally. The filter always passes through the first
19088 input. The signature of each stream can be written into a file.
19089 It accepts the following options:
19091 detectmode
19093 Enable or disable the matching process.
19094 Available values are:
19096 off Disable the calculation of a matching (default).
19098 full
19100 Calculate the matching for the whole video and output whether
19101 the whole video matches or only parts.
19102 fast
19104 Calculate only until a matching is found or the video ends.
19105 Should be faster in some cases.
19106 nb_inputs
19108 Set the number of inputs. The option value must be a non negative
19109 integer. Default value is 1.
19110 filename
19112 Set the path to which the output is written. If there is more than
19113 one input, the path must be a prototype, i.e. must contain %d or
19114 %0nd (where n is a positive integer), that will be replaced with
19115 the input number. If no filename is specified, no output will be
19116 written. This is the default.
19117 format
19119 Choose the output format.
19120 Available values are:
19122 binary
19124 Use the specified binary representation (default).
19125 xml Use the specified xml representation.
19127 th_d
19129 Set threshold to detect one word as similar. The option value must
19130 be an integer greater than zero. The default value is 9000.
19131 th_dc
19133 Set threshold to detect all words as similar. The option value must
19134 be an integer greater than zero. The default value is 60000.
19135 th_xh
19137 Set threshold to detect frames as similar. The option value must be
19138 an integer greater than zero. The default value is 116.
19139 th_di
19141 Set the minimum length of a sequence in frames to recognize it as
19142 matching sequence. The option value must be a non negative integer
19143 value. The default value is 0.
19144 th_it
19146 Set the minimum relation, that matching frames to all frames must
19147 have. The option value must be a double value between 0 and 1. The
19148 default value is 0.5.
19149 Examples
19151 • To calculate the signature of an input video and store it in
19153 signature.bin:
19154 ffmpeg -i input.mkv -vf signature=filename=signature.bin -map 0:v -f null -
19156 • To detect whether two videos match and store the signatures in XML
19158 format in signature0.xml and signature1.xml:
19159 ffmpeg -i input1.mkv -i input2.mkv -filter_complex "[0:v][1:v] signature=nb_inputs=2:detectmode=full:format=xml:filename=signature%d.xml" -map :v -f null -
19161 siti
19163 Calculate Spatial Info (SI) and Temporal Info (TI) scores for a video,
19164 as defined in ITU-T P.910: Subjective video quality assessment methods
19165 for multimedia applications. Available PDF at
19166 <https://www.itu.int/rec/T-REC-P.910-199909-S/en >.
19167 It accepts the following option:
19169 print_summary
19171 If set to 1, Summary statistics will be printed to the console.
19172 Default 0.
19173 Examples
19175 • To calculate SI/TI metrics and print summary:
19177 ffmpeg -i input.mp4 -vf siti=print_summary=1 -f null -
19179 smartblur
19181 Blur the input video without impacting the outlines.
19182 It accepts the following options:
19184 luma_radius, lr
19186 Set the luma radius. The option value must be a float number in the
19187 range [0.1,5.0] that specifies the variance of the gaussian filter
19188 used to blur the image (slower if larger). Default value is 1.0.
19189 luma_strength, ls
19191 Set the luma strength. The option value must be a float number in
19192 the range [-1.0,1.0] that configures the blurring. A value included
19193 in [0.0,1.0] will blur the image whereas a value included in
19194 [-1.0,0.0] will sharpen the image. Default value is 1.0.
19195 luma_threshold, lt
19197 Set the luma threshold used as a coefficient to determine whether a
19198 pixel should be blurred or not. The option value must be an integer
19199 in the range [-30,30]. A value of 0 will filter all the image, a
19200 value included in [0,30] will filter flat areas and a value
19201 included in [-30,0] will filter edges. Default value is 0.
19202 chroma_radius, cr
19204 Set the chroma radius. The option value must be a float number in
19205 the range [0.1,5.0] that specifies the variance of the gaussian
19206 filter used to blur the image (slower if larger). Default value is
19207 luma_radius.
19208 chroma_strength, cs
19210 Set the chroma strength. The option value must be a float number in
19211 the range [-1.0,1.0] that configures the blurring. A value included
19212 in [0.0,1.0] will blur the image whereas a value included in
19213 [-1.0,0.0] will sharpen the image. Default value is luma_strength.
19214 chroma_threshold, ct
19216 Set the chroma threshold used as a coefficient to determine whether
19217 a pixel should be blurred or not. The option value must be an
19218 integer in the range [-30,30]. A value of 0 will filter all the
19219 image, a value included in [0,30] will filter flat areas and a
19220 value included in [-30,0] will filter edges. Default value is
19221 luma_threshold.
19222 If a chroma option is not explicitly set, the corresponding luma value
19224 is set.
19225 sobel
19227 Apply sobel operator to input video stream.
19228 The filter accepts the following option:
19230 planes
19232 Set which planes will be processed, unprocessed planes will be
19233 copied. By default value 0xf, all planes will be processed.
19234 scale
19236 Set value which will be multiplied with filtered result.
19237 delta
19239 Set value which will be added to filtered result.
19240 Commands
19242 This filter supports the all above options as commands.
19244 spp
19246 Apply a simple postprocessing filter that compresses and decompresses
19247 the image at several (or - in the case of quality level 6 - all) shifts
19248 and average the results.
19249 The filter accepts the following options:
19251 quality
19253 Set quality. This option defines the number of levels for
19254 averaging. It accepts an integer in the range 0-6. If set to 0, the
19255 filter will have no effect. A value of 6 means the higher quality.
19256 For each increment of that value the speed drops by a factor of
19257 approximately 2. Default value is 3.
19258 qp Force a constant quantization parameter. If not set, the filter
19260 will use the QP from the video stream (if available).
19261 mode
19263 Set thresholding mode. Available modes are:
19264 hard
19266 Set hard thresholding (default).
19267 soft
19269 Set soft thresholding (better de-ringing effect, but likely
19270 blurrier).
19271 use_bframe_qp
19273 Enable the use of the QP from the B-Frames if set to 1. Using this
19274 option may cause flicker since the B-Frames have often larger QP.
19275 Default is 0 (not enabled).
19276 Commands
19278 This filter supports the following commands:
19280 quality, level
19282 Set quality level. The value "max" can be used to set the maximum
19283 level, currently 6.
19284 sr
19286 Scale the input by applying one of the super-resolution methods based
19287 on convolutional neural networks. Supported models:
19288 • Super-Resolution Convolutional Neural Network model (SRCNN). See
19290 <https://arxiv.org/abs/1501.00092>.
19291 • Efficient Sub-Pixel Convolutional Neural Network model (ESPCN).
19293 See <https://arxiv.org/abs/1609.05158>.
19294 Training scripts as well as scripts for model file (.pb) saving can be
19296 found at <https://github.com/XueweiMeng/sr/tree/sr_dnn_native>.
19297 Original repository is at
19298 <https://github.com/HighVoltageRocknRoll/sr.git>.
19299 Native model files (.model) can be generated from TensorFlow model
19301 files (.pb) by using tools/python/convert.py
19302 The filter accepts the following options:
19304 dnn_backend
19306 Specify which DNN backend to use for model loading and execution.
19307 This option accepts the following values:
19308 native
19310 Native implementation of DNN loading and execution.
19311 tensorflow
19313 TensorFlow backend. To enable this backend you need to install
19314 the TensorFlow for C library (see
19315 <https://www.tensorflow.org/install/lang_c>) and configure
19316 FFmpeg with "--enable-libtensorflow"
19317 Default value is native.
19319 model
19321 Set path to model file specifying network architecture and its
19322 parameters. Note that different backends use different file
19323 formats. TensorFlow backend can load files for both formats, while
19324 native backend can load files for only its format.
19325 scale_factor
19327 Set scale factor for SRCNN model. Allowed values are 2, 3 and 4.
19328 Default value is 2. Scale factor is necessary for SRCNN model,
19329 because it accepts input upscaled using bicubic upscaling with
19330 proper scale factor.
19331 To get full functionality (such as async execution), please use the
19333 dnn_processing filter.
19334 ssim
19336 Obtain the SSIM (Structural SImilarity Metric) between two input
19337 videos.
19338 This filter takes in input two input videos, the first input is
19340 considered the "main" source and is passed unchanged to the output. The
19341 second input is used as a "reference" video for computing the SSIM.
19342 Both video inputs must have the same resolution and pixel format for
19344 this filter to work correctly. Also it assumes that both inputs have
19345 the same number of frames, which are compared one by one.
19346 The filter stores the calculated SSIM of each frame.
19348 The description of the accepted parameters follows.
19350 stats_file, f
19352 If specified the filter will use the named file to save the SSIM of
19353 each individual frame. When filename equals "-" the data is sent to
19354 standard output.
19355 The file printed if stats_file is selected, contains a sequence of
19357 key/value pairs of the form key:value for each compared couple of
19358 frames.
19359 A description of each shown parameter follows:
19361 n sequential number of the input frame, starting from 1
19363 Y, U, V, R, G, B
19365 SSIM of the compared frames for the component specified by the
19366 suffix.
19367 All SSIM of the compared frames for the whole frame.
19369 dB Same as above but in dB representation.
19371 This filter also supports the framesync options.
19373 Examples
19375 • For example:
19377 movie=ref_movie.mpg, setpts=PTS-STARTPTS [main];
19379 [main][ref] ssim="stats_file=stats.log" [out]
19380 On this example the input file being processed is compared with the
19382 reference file ref_movie.mpg. The SSIM of each individual frame is
19383 stored in stats.log.
19384 • Another example with both psnr and ssim at same time:
19386 ffmpeg -i main.mpg -i ref.mpg -lavfi "ssim;[0:v][1:v]psnr" -f null -
19388 • Another example with different containers:
19390 ffmpeg -i main.mpg -i ref.mkv -lavfi "[0:v]settb=AVTB,setpts=PTS-STARTPTS[main];[1:v]settb=AVTB,setpts=PTS-STARTPTS[ref];[main][ref]ssim" -f null -
19392 stereo3d
19394 Convert between different stereoscopic image formats.
19395 The filters accept the following options:
19397 in Set stereoscopic image format of input.
19399 Available values for input image formats are:
19401 sbsl
19403 side by side parallel (left eye left, right eye right)
19404 sbsr
19406 side by side crosseye (right eye left, left eye right)
19407 sbs2l
19409 side by side parallel with half width resolution (left eye
19410 left, right eye right)
19411 sbs2r
19413 side by side crosseye with half width resolution (right eye
19414 left, left eye right)
19415 abl
19417 tbl above-below (left eye above, right eye below)
19418 abr
19420 tbr above-below (right eye above, left eye below)
19421 ab2l
19423 tb2l
19424 above-below with half height resolution (left eye above, right
19425 eye below)
19426 ab2r
19428 tb2r
19429 above-below with half height resolution (right eye above, left
19430 eye below)
19431 al alternating frames (left eye first, right eye second)
19433 ar alternating frames (right eye first, left eye second)
19435 irl interleaved rows (left eye has top row, right eye starts on
19437 next row)
19438 irr interleaved rows (right eye has top row, left eye starts on
19440 next row)
19441 icl interleaved columns, left eye first
19443 icr interleaved columns, right eye first
19445 Default value is sbsl.
19447 out Set stereoscopic image format of output.
19449 sbsl
19451 side by side parallel (left eye left, right eye right)
19452 sbsr
19454 side by side crosseye (right eye left, left eye right)
19455 sbs2l
19457 side by side parallel with half width resolution (left eye
19458 left, right eye right)
19459 sbs2r
19461 side by side crosseye with half width resolution (right eye
19462 left, left eye right)
19463 abl
19465 tbl above-below (left eye above, right eye below)
19466 abr
19468 tbr above-below (right eye above, left eye below)
19469 ab2l
19471 tb2l
19472 above-below with half height resolution (left eye above, right
19473 eye below)
19474 ab2r
19476 tb2r
19477 above-below with half height resolution (right eye above, left
19478 eye below)
19479 al alternating frames (left eye first, right eye second)
19481 ar alternating frames (right eye first, left eye second)
19483 irl interleaved rows (left eye has top row, right eye starts on
19485 next row)
19486 irr interleaved rows (right eye has top row, left eye starts on
19488 next row)
19489 arbg
19491 anaglyph red/blue gray (red filter on left eye, blue filter on
19492 right eye)
19493 argg
19495 anaglyph red/green gray (red filter on left eye, green filter
19496 on right eye)
19497 arcg
19499 anaglyph red/cyan gray (red filter on left eye, cyan filter on
19500 right eye)
19501 arch
19503 anaglyph red/cyan half colored (red filter on left eye, cyan
19504 filter on right eye)
19505 arcc
19507 anaglyph red/cyan color (red filter on left eye, cyan filter on
19508 right eye)
19509 arcd
19511 anaglyph red/cyan color optimized with the least squares
19512 projection of dubois (red filter on left eye, cyan filter on
19513 right eye)
19514 agmg
19516 anaglyph green/magenta gray (green filter on left eye, magenta
19517 filter on right eye)
19518 agmh
19520 anaglyph green/magenta half colored (green filter on left eye,
19521 magenta filter on right eye)
19522 agmc
19524 anaglyph green/magenta colored (green filter on left eye,
19525 magenta filter on right eye)
19526 agmd
19528 anaglyph green/magenta color optimized with the least squares
19529 projection of dubois (green filter on left eye, magenta filter
19530 on right eye)
19531 aybg
19533 anaglyph yellow/blue gray (yellow filter on left eye, blue
19534 filter on right eye)
19535 aybh
19537 anaglyph yellow/blue half colored (yellow filter on left eye,
19538 blue filter on right eye)
19539 aybc
19541 anaglyph yellow/blue colored (yellow filter on left eye, blue
19542 filter on right eye)
19543 aybd
19545 anaglyph yellow/blue color optimized with the least squares
19546 projection of dubois (yellow filter on left eye, blue filter on
19547 right eye)
19548 ml mono output (left eye only)
19550 mr mono output (right eye only)
19552 chl checkerboard, left eye first
19554 chr checkerboard, right eye first
19556 icl interleaved columns, left eye first
19558 icr interleaved columns, right eye first
19560 hdmi
19562 HDMI frame pack
19563 Default value is arcd.
19565 Examples
19567 • Convert input video from side by side parallel to anaglyph
19569 yellow/blue dubois:
19570 stereo3d=sbsl:aybd
19572 • Convert input video from above below (left eye above, right eye
19574 below) to side by side crosseye.
19575 stereo3d=abl:sbsr
19577 streamselect, astreamselect
19579 Select video or audio streams.
19580 The filter accepts the following options:
19582 inputs
19584 Set number of inputs. Default is 2.
19585 map Set input indexes to remap to outputs.
19587 Commands
19589 The "streamselect" and "astreamselect" filter supports the following
19591 commands:
19592 map Set input indexes to remap to outputs.
19594 Examples
19596 • Select first 5 seconds 1st stream and rest of time 2nd stream:
19598 sendcmd='5.0 streamselect map 1',streamselect=inputs=2:map=0
19600 • Same as above, but for audio:
19602 asendcmd='5.0 astreamselect map 1',astreamselect=inputs=2:map=0
19604 subtitles
19606 Draw subtitles on top of input video using the libass library.
19607 To enable compilation of this filter you need to configure FFmpeg with
19609 "--enable-libass". This filter also requires a build with libavcodec
19610 and libavformat to convert the passed subtitles file to ASS (Advanced
19611 Substation Alpha) subtitles format.
19612 The filter accepts the following options:
19614 filename, f
19616 Set the filename of the subtitle file to read. It must be
19617 specified.
19618 original_size
19620 Specify the size of the original video, the video for which the ASS
19621 file was composed. For the syntax of this option, check the "Video
19622 size" section in the ffmpeg-utils manual. Due to a misdesign in
19623 ASS aspect ratio arithmetic, this is necessary to correctly scale
19624 the fonts if the aspect ratio has been changed.
19625 fontsdir
19627 Set a directory path containing fonts that can be used by the
19628 filter. These fonts will be used in addition to whatever the font
19629 provider uses.
19630 alpha
19632 Process alpha channel, by default alpha channel is untouched.
19633 charenc
19635 Set subtitles input character encoding. "subtitles" filter only.
19636 Only useful if not UTF-8.
19637 stream_index, si
19639 Set subtitles stream index. "subtitles" filter only.
19640 force_style
19642 Override default style or script info parameters of the subtitles.
19643 It accepts a string containing ASS style format "KEY=VALUE" couples
19644 separated by ",".
19645 If the first key is not specified, it is assumed that the first value
19647 specifies the filename.
19648 For example, to render the file sub.srt on top of the input video, use
19650 the command:
19651 subtitles=sub.srt
19653 which is equivalent to:
19655 subtitles=filename=sub.srt
19657 To render the default subtitles stream from file video.mkv, use:
19659 subtitles=video.mkv
19661 To render the second subtitles stream from that file, use:
19663 subtitles=video.mkv:si=1
19665 To make the subtitles stream from sub.srt appear in 80% transparent
19667 blue "DejaVu Serif", use:
19668 subtitles=sub.srt:force_style='Fontname=DejaVu Serif,PrimaryColour=&HCCFF0000'
19670 super2xsai
19672 Scale the input by 2x and smooth using the Super2xSaI (Scale and
19673 Interpolate) pixel art scaling algorithm.
19674 Useful for enlarging pixel art images without reducing sharpness.
19676 swaprect
19678 Swap two rectangular objects in video.
19679 This filter accepts the following options:
19681 w Set object width.
19683 h Set object height.
19685 x1 Set 1st rect x coordinate.
19687 y1 Set 1st rect y coordinate.
19689 x2 Set 2nd rect x coordinate.
19691 y2 Set 2nd rect y coordinate.
19693 All expressions are evaluated once for each frame.
19695 The all options are expressions containing the following constants:
19697 w
19699 h The input width and height.
19700 a same as w / h
19702 sar input sample aspect ratio
19704 dar input display aspect ratio, it is the same as (w / h) * sar
19706 n The number of the input frame, starting from 0.
19708 t The timestamp expressed in seconds. It's NAN if the input timestamp
19710 is unknown.
19711 pos the position in the file of the input frame, NAN if unknown
19713 Commands
19715 This filter supports the all above options as commands.
19717 swapuv
19719 Swap U & V plane.
19720 tblend
19722 Blend successive video frames.
19723 See blend
19725 telecine
19727 Apply telecine process to the video.
19728 This filter accepts the following options:
19730 first_field
19732 top, t
19733 top field first
19734 bottom, b
19736 bottom field first The default value is "top".
19737 pattern
19739 A string of numbers representing the pulldown pattern you wish to
19740 apply. The default value is 23.
19741 Some typical patterns:
19743 NTSC output (30i):
19745 27.5p: 32222
19746 24p: 23 (classic)
19747 24p: 2332 (preferred)
19748 20p: 33
19749 18p: 334
19750 16p: 3444
19751 PAL output (25i):
19753 27.5p: 12222
19754 24p: 222222222223 ("Euro pulldown")
19755 16.67p: 33
19756 16p: 33333334
19757 thistogram
19759 Compute and draw a color distribution histogram for the input video
19760 across time.
19761 Unlike histogram video filter which only shows histogram of single
19763 input frame at certain time, this filter shows also past histograms of
19764 number of frames defined by "width" option.
19765 The computed histogram is a representation of the color component
19767 distribution in an image.
19768 The filter accepts the following options:
19770 width, w
19772 Set width of single color component output. Default value is 0.
19773 Value of 0 means width will be picked from input video. This also
19774 set number of passed histograms to keep. Allowed range is [0,
19775 8192].
19776 display_mode, d
19778 Set display mode. It accepts the following values:
19779 stack
19781 Per color component graphs are placed below each other.
19782 parade
19784 Per color component graphs are placed side by side.
19785 overlay
19787 Presents information identical to that in the "parade", except
19788 that the graphs representing color components are superimposed
19789 directly over one another.
19790 Default is "stack".
19792 levels_mode, m
19794 Set mode. Can be either "linear", or "logarithmic". Default is
19795 "linear".
19796 components, c
19798 Set what color components to display. Default is 7.
19799 bgopacity, b
19801 Set background opacity. Default is 0.9.
19802 envelope, e
19804 Show envelope. Default is disabled.
19805 ecolor, ec
19807 Set envelope color. Default is "gold".
19808 slide
19810 Set slide mode.
19811 Available values for slide is:
19813 frame
19815 Draw new frame when right border is reached.
19816 replace
19818 Replace old columns with new ones.
19819 scroll
19821 Scroll from right to left.
19822 rscroll
19824 Scroll from left to right.
19825 picture
19827 Draw single picture.
19828 Default is "replace".
19830 threshold
19832 Apply threshold effect to video stream.
19833 This filter needs four video streams to perform thresholding. First
19835 stream is stream we are filtering. Second stream is holding threshold
19836 values, third stream is holding min values, and last, fourth stream is
19837 holding max values.
19838 The filter accepts the following option:
19840 planes
19842 Set which planes will be processed, unprocessed planes will be
19843 copied. By default value 0xf, all planes will be processed.
19844 For example if first stream pixel's component value is less then
19846 threshold value of pixel component from 2nd threshold stream, third
19847 stream value will picked, otherwise fourth stream pixel component value
19848 will be picked.
19849 Using color source filter one can perform various types of
19851 thresholding:
19852 Commands
19854 This filter supports the all options as commands.
19856 Examples
19858 • Binary threshold, using gray color as threshold:
19860 ffmpeg -i 320x240.avi -f lavfi -i color=gray -f lavfi -i color=black -f lavfi -i color=white -lavfi threshold output.avi
19862 • Inverted binary threshold, using gray color as threshold:
19864 ffmpeg -i 320x240.avi -f lavfi -i color=gray -f lavfi -i color=white -f lavfi -i color=black -lavfi threshold output.avi
19866 • Truncate binary threshold, using gray color as threshold:
19868 ffmpeg -i 320x240.avi -f lavfi -i color=gray -i 320x240.avi -f lavfi -i color=gray -lavfi threshold output.avi
19870 • Threshold to zero, using gray color as threshold:
19872 ffmpeg -i 320x240.avi -f lavfi -i color=gray -f lavfi -i color=white -i 320x240.avi -lavfi threshold output.avi
19874 • Inverted threshold to zero, using gray color as threshold:
19876 ffmpeg -i 320x240.avi -f lavfi -i color=gray -i 320x240.avi -f lavfi -i color=white -lavfi threshold output.avi
19878 thumbnail
19880 Select the most representative frame in a given sequence of consecutive
19881 frames.
19882 The filter accepts the following options:
19884 n Set the frames batch size to analyze; in a set of n frames, the
19886 filter will pick one of them, and then handle the next batch of n
19887 frames until the end. Default is 100.
19888 log Set the log level to display picked frame stats. Default is
19890 "info".
19891 Since the filter keeps track of the whole frames sequence, a bigger n
19893 value will result in a higher memory usage, so a high value is not
19894 recommended.
19895 Examples
19897 • Extract one picture each 50 frames:
19899 thumbnail=50
19901 • Complete example of a thumbnail creation with ffmpeg:
19903 ffmpeg -i in.avi -vf thumbnail,scale=300:200 -frames:v 1 out.png
19905 tile
19907 Tile several successive frames together.
19908 The untile filter can do the reverse.
19910 The filter accepts the following options:
19912 layout
19914 Set the grid size in the form "COLUMNSxROWS". Range is upto
19915 UINT_MAX cells. Default is "6x5".
19916 nb_frames
19918 Set the maximum number of frames to render in the given area. It
19919 must be less than or equal to wxh. The default value is 0, meaning
19920 all the area will be used.
19921 margin
19923 Set the outer border margin in pixels. Range is 0 to 1024. Default
19924 is 0.
19925 padding
19927 Set the inner border thickness (i.e. the number of pixels between
19928 frames). For more advanced padding options (such as having
19929 different values for the edges), refer to the pad video filter.
19930 Range is 0 to 1024. Default is 0.
19931 color
19933 Specify the color of the unused area. For the syntax of this
19934 option, check the "Color" section in the ffmpeg-utils manual. The
19935 default value of color is "black".
19936 overlap
19938 Set the number of frames to overlap when tiling several successive
19939 frames together. The value must be between 0 and nb_frames - 1.
19940 Default is 0.
19941 init_padding
19943 Set the number of frames to initially be empty before displaying
19944 first output frame. This controls how soon will one get first
19945 output frame. The value must be between 0 and nb_frames - 1.
19946 Default is 0.
19947 Examples
19949 • Produce 8x8 PNG tiles of all keyframes (-skip_frame nokey) in a
19951 movie:
19952 ffmpeg -skip_frame nokey -i file.avi -vf 'scale=128:72,tile=8x8' -an -vsync 0 keyframes%03d.png
19954 The -vsync 0 is necessary to prevent ffmpeg from duplicating each
19956 output frame to accommodate the originally detected frame rate.
19957 • Display 5 pictures in an area of "3x2" frames, with 7 pixels
19959 between them, and 2 pixels of initial margin, using mixed flat and
19960 named options:
19961 tile=3x2:nb_frames=5:padding=7:margin=2
19963 tinterlace
19965 Perform various types of temporal field interlacing.
19966 Frames are counted starting from 1, so the first input frame is
19968 considered odd.
19969 The filter accepts the following options:
19971 mode
19973 Specify the mode of the interlacing. This option can also be
19974 specified as a value alone. See below for a list of values for this
19975 option.
19976 Available values are:
19978 merge, 0
19980 Move odd frames into the upper field, even into the lower
19981 field, generating a double height frame at half frame rate.
19982 ------> time
19984 Input:
19985 Frame 1 Frame 2 Frame 3 Frame 4
19986 11111 22222 33333 44444
19988 11111 22222 33333 44444
19989 11111 22222 33333 44444
19990 11111 22222 33333 44444
19991 Output:
19993 11111 33333
19994 22222 44444
19995 11111 33333
19996 22222 44444
19997 11111 33333
19998 22222 44444
19999 11111 33333
20000 22222 44444
20001 drop_even, 1
20003 Only output odd frames, even frames are dropped, generating a
20004 frame with unchanged height at half frame rate.
20005 ------> time
20007 Input:
20008 Frame 1 Frame 2 Frame 3 Frame 4
20009 11111 22222 33333 44444
20011 11111 22222 33333 44444
20012 11111 22222 33333 44444
20013 11111 22222 33333 44444
20014 Output:
20016 11111 33333
20017 11111 33333
20018 11111 33333
20019 11111 33333
20020 drop_odd, 2
20022 Only output even frames, odd frames are dropped, generating a
20023 frame with unchanged height at half frame rate.
20024 ------> time
20026 Input:
20027 Frame 1 Frame 2 Frame 3 Frame 4
20028 11111 22222 33333 44444
20030 11111 22222 33333 44444
20031 11111 22222 33333 44444
20032 11111 22222 33333 44444
20033 Output:
20035 22222 44444
20036 22222 44444
20037 22222 44444
20038 22222 44444
20039 pad, 3
20041 Expand each frame to full height, but pad alternate lines with
20042 black, generating a frame with double height at the same input
20043 frame rate.
20044 ------> time
20046 Input:
20047 Frame 1 Frame 2 Frame 3 Frame 4
20048 11111 22222 33333 44444
20050 11111 22222 33333 44444
20051 11111 22222 33333 44444
20052 11111 22222 33333 44444
20053 Output:
20055 11111 ..... 33333 .....
20056 ..... 22222 ..... 44444
20057 11111 ..... 33333 .....
20058 ..... 22222 ..... 44444
20059 11111 ..... 33333 .....
20060 ..... 22222 ..... 44444
20061 11111 ..... 33333 .....
20062 ..... 22222 ..... 44444
20063 interleave_top, 4
20065 Interleave the upper field from odd frames with the lower field
20066 from even frames, generating a frame with unchanged height at
20067 half frame rate.
20068 ------> time
20070 Input:
20071 Frame 1 Frame 2 Frame 3 Frame 4
20072 11111<- 22222 33333<- 44444
20074 11111 22222<- 33333 44444<-
20075 11111<- 22222 33333<- 44444
20076 11111 22222<- 33333 44444<-
20077 Output:
20079 11111 33333
20080 22222 44444
20081 11111 33333
20082 22222 44444
20083 interleave_bottom, 5
20085 Interleave the lower field from odd frames with the upper field
20086 from even frames, generating a frame with unchanged height at
20087 half frame rate.
20088 ------> time
20090 Input:
20091 Frame 1 Frame 2 Frame 3 Frame 4
20092 11111 22222<- 33333 44444<-
20094 11111<- 22222 33333<- 44444
20095 11111 22222<- 33333 44444<-
20096 11111<- 22222 33333<- 44444
20097 Output:
20099 22222 44444
20100 11111 33333
20101 22222 44444
20102 11111 33333
20103 interlacex2, 6
20105 Double frame rate with unchanged height. Frames are inserted
20106 each containing the second temporal field from the previous
20107 input frame and the first temporal field from the next input
20108 frame. This mode relies on the top_field_first flag. Useful for
20109 interlaced video displays with no field synchronisation.
20110 ------> time
20112 Input:
20113 Frame 1 Frame 2 Frame 3 Frame 4
20114 11111 22222 33333 44444
20116 11111 22222 33333 44444
20117 11111 22222 33333 44444
20118 11111 22222 33333 44444
20119 Output:
20121 11111 22222 22222 33333 33333 44444 44444
20122 11111 11111 22222 22222 33333 33333 44444
20123 11111 22222 22222 33333 33333 44444 44444
20124 11111 11111 22222 22222 33333 33333 44444
20125 mergex2, 7
20127 Move odd frames into the upper field, even into the lower
20128 field, generating a double height frame at same frame rate.
20129 ------> time
20131 Input:
20132 Frame 1 Frame 2 Frame 3 Frame 4
20133 11111 22222 33333 44444
20135 11111 22222 33333 44444
20136 11111 22222 33333 44444
20137 11111 22222 33333 44444
20138 Output:
20140 11111 33333 33333 55555
20141 22222 22222 44444 44444
20142 11111 33333 33333 55555
20143 22222 22222 44444 44444
20144 11111 33333 33333 55555
20145 22222 22222 44444 44444
20146 11111 33333 33333 55555
20147 22222 22222 44444 44444
20148 Numeric values are deprecated but are accepted for backward
20150 compatibility reasons.
20151 Default mode is "merge".
20153 flags
20155 Specify flags influencing the filter process.
20156 Available value for flags is:
20158 low_pass_filter, vlpf
20160 Enable linear vertical low-pass filtering in the filter.
20161 Vertical low-pass filtering is required when creating an
20162 interlaced destination from a progressive source which contains
20163 high-frequency vertical detail. Filtering will reduce interlace
20164 'twitter' and Moire patterning.
20165 complex_filter, cvlpf
20167 Enable complex vertical low-pass filtering. This will slightly
20168 less reduce interlace 'twitter' and Moire patterning but better
20169 retain detail and subjective sharpness impression.
20170 bypass_il
20172 Bypass already interlaced frames, only adjust the frame rate.
20173 Vertical low-pass filtering and bypassing already interlaced frames
20175 can only be enabled for mode interleave_top and interleave_bottom.
20176 tmedian
20178 Pick median pixels from several successive input video frames.
20179 The filter accepts the following options:
20181 radius
20183 Set radius of median filter. Default is 1. Allowed range is from 1
20184 to 127.
20185 planes
20187 Set which planes to filter. Default value is 15, by which all
20188 planes are processed.
20189 percentile
20191 Set median percentile. Default value is 0.5. Default value of 0.5
20192 will pick always median values, while 0 will pick minimum values,
20193 and 1 maximum values.
20194 Commands
20196 This filter supports all above options as commands, excluding option
20198 "radius".
20199 tmidequalizer
20201 Apply Temporal Midway Video Equalization effect.
20202 Midway Video Equalization adjusts a sequence of video frames to have
20204 the same histograms, while maintaining their dynamics as much as
20205 possible. It's useful for e.g. matching exposures from a video frames
20206 sequence.
20207 This filter accepts the following option:
20209 radius
20211 Set filtering radius. Default is 5. Allowed range is from 1 to 127.
20212 sigma
20214 Set filtering sigma. Default is 0.5. This controls strength of
20215 filtering. Setting this option to 0 effectively does nothing.
20216 planes
20218 Set which planes to process. Default is 15, which is all available
20219 planes.
20220 tmix
20222 Mix successive video frames.
20223 A description of the accepted options follows.
20225 frames
20227 The number of successive frames to mix. If unspecified, it defaults
20228 to 3.
20229 weights
20231 Specify weight of each input video frame. Each weight is separated
20232 by space. If number of weights is smaller than number of frames
20233 last specified weight will be used for all remaining unset weights.
20234 scale
20236 Specify scale, if it is set it will be multiplied with sum of each
20237 weight multiplied with pixel values to give final destination pixel
20238 value. By default scale is auto scaled to sum of weights.
20239 planes
20241 Set which planes to filter. Default is all. Allowed range is from 0
20242 to 15.
20243 Examples
20245 • Average 7 successive frames:
20247 tmix=frames=7:weights="1 1 1 1 1 1 1"
20249 • Apply simple temporal convolution:
20251 tmix=frames=3:weights="-1 3 -1"
20253 • Similar as above but only showing temporal differences:
20255 tmix=frames=3:weights="-1 2 -1":scale=1
20257 Commands
20259 This filter supports the following commands:
20261 weights
20263 scale
20264 planes
20265 Syntax is same as option with same name.
20266 tonemap
20268 Tone map colors from different dynamic ranges.
20269 This filter expects data in single precision floating point, as it
20271 needs to operate on (and can output) out-of-range values. Another
20272 filter, such as zscale, is needed to convert the resulting frame to a
20273 usable format.
20274 The tonemapping algorithms implemented only work on linear light, so
20276 input data should be linearized beforehand (and possibly correctly
20277 tagged).
20278 ffmpeg -i INPUT -vf zscale=transfer=linear,tonemap=clip,zscale=transfer=bt709,format=yuv420p OUTPUT
20280 Options
20282 The filter accepts the following options.
20284 tonemap
20286 Set the tone map algorithm to use.
20287 Possible values are:
20289 none
20291 Do not apply any tone map, only desaturate overbright pixels.
20292 clip
20294 Hard-clip any out-of-range values. Use it for perfect color
20295 accuracy for in-range values, while distorting out-of-range
20296 values.
20297 linear
20299 Stretch the entire reference gamut to a linear multiple of the
20300 display.
20301 gamma
20303 Fit a logarithmic transfer between the tone curves.
20304 reinhard
20306 Preserve overall image brightness with a simple curve, using
20307 nonlinear contrast, which results in flattening details and
20308 degrading color accuracy.
20309 hable
20311 Preserve both dark and bright details better than reinhard, at
20312 the cost of slightly darkening everything. Use it when detail
20313 preservation is more important than color and brightness
20314 accuracy.
20315 mobius
20317 Smoothly map out-of-range values, while retaining contrast and
20318 colors for in-range material as much as possible. Use it when
20319 color accuracy is more important than detail preservation.
20320 Default is none.
20322 param
20324 Tune the tone mapping algorithm.
20325 This affects the following algorithms:
20327 none
20329 Ignored.
20330 linear
20332 Specifies the scale factor to use while stretching. Default to
20333 1.0.
20334 gamma
20336 Specifies the exponent of the function. Default to 1.8.
20337 clip
20339 Specify an extra linear coefficient to multiply into the signal
20340 before clipping. Default to 1.0.
20341 reinhard
20343 Specify the local contrast coefficient at the display peak.
20344 Default to 0.5, which means that in-gamut values will be about
20345 half as bright as when clipping.
20346 hable
20348 Ignored.
20349 mobius
20351 Specify the transition point from linear to mobius transform.
20352 Every value below this point is guaranteed to be mapped 1:1.
20353 The higher the value, the more accurate the result will be, at
20354 the cost of losing bright details. Default to 0.3, which due
20355 to the steep initial slope still preserves in-range colors
20356 fairly accurately.
20357 desat
20359 Apply desaturation for highlights that exceed this level of
20360 brightness. The higher the parameter, the more color information
20361 will be preserved. This setting helps prevent unnaturally blown-out
20362 colors for super-highlights, by (smoothly) turning into white
20363 instead. This makes images feel more natural, at the cost of
20364 reducing information about out-of-range colors.
20365 The default of 2.0 is somewhat conservative and will mostly just
20367 apply to skies or directly sunlit surfaces. A setting of 0.0
20368 disables this option.
20369 This option works only if the input frame has a supported color
20371 tag.
20372 peak
20374 Override signal/nominal/reference peak with this value. Useful when
20375 the embedded peak information in display metadata is not reliable
20376 or when tone mapping from a lower range to a higher range.
20377 tpad
20379 Temporarily pad video frames.
20380 The filter accepts the following options:
20382 start
20384 Specify number of delay frames before input video stream. Default
20385 is 0.
20386 stop
20388 Specify number of padding frames after input video stream. Set to
20389 -1 to pad indefinitely. Default is 0.
20390 start_mode
20392 Set kind of frames added to beginning of stream. Can be either add
20393 or clone. With add frames of solid-color are added. With clone
20394 frames are clones of first frame. Default is add.
20395 stop_mode
20397 Set kind of frames added to end of stream. Can be either add or
20398 clone. With add frames of solid-color are added. With clone
20399 frames are clones of last frame. Default is add.
20400 start_duration, stop_duration
20402 Specify the duration of the start/stop delay. See the Time duration
20403 section in the ffmpeg-utils(1) manual for the accepted syntax.
20404 These options override start and stop. Default is 0.
20405 color
20407 Specify the color of the padded area. For the syntax of this
20408 option, check the "Color" section in the ffmpeg-utils manual.
20409 The default value of color is "black".
20411 transpose
20413 Transpose rows with columns in the input video and optionally flip it.
20414 It accepts the following parameters:
20416 dir Specify the transposition direction.
20418 Can assume the following values:
20420 0, 4, cclock_flip
20422 Rotate by 90 degrees counterclockwise and vertically flip
20423 (default), that is:
20424 L.R L.l
20426 . . -> . .
20427 l.r R.r
20428 1, 5, clock
20430 Rotate by 90 degrees clockwise, that is:
20431 L.R l.L
20433 . . -> . .
20434 l.r r.R
20435 2, 6, cclock
20437 Rotate by 90 degrees counterclockwise, that is:
20438 L.R R.r
20440 . . -> . .
20441 l.r L.l
20442 3, 7, clock_flip
20444 Rotate by 90 degrees clockwise and vertically flip, that is:
20445 L.R r.R
20447 . . -> . .
20448 l.r l.L
20449 For values between 4-7, the transposition is only done if the input
20451 video geometry is portrait and not landscape. These values are
20452 deprecated, the "passthrough" option should be used instead.
20453 Numerical values are deprecated, and should be dropped in favor of
20455 symbolic constants.
20456 passthrough
20458 Do not apply the transposition if the input geometry matches the
20459 one specified by the specified value. It accepts the following
20460 values:
20461 none
20463 Always apply transposition.
20464 portrait
20466 Preserve portrait geometry (when height >= width).
20467 landscape
20469 Preserve landscape geometry (when width >= height).
20470 Default value is "none".
20472 For example to rotate by 90 degrees clockwise and preserve portrait
20474 layout:
20475 transpose=dir=1:passthrough=portrait
20477 The command above can also be specified as:
20479 transpose=1:portrait
20481 transpose_npp
20483 Transpose rows with columns in the input video and optionally flip it.
20484 For more in depth examples see the transpose video filter, which shares
20485 mostly the same options.
20486 It accepts the following parameters:
20488 dir Specify the transposition direction.
20490 Can assume the following values:
20492 cclock_flip
20494 Rotate by 90 degrees counterclockwise and vertically flip.
20495 (default)
20496 clock
20498 Rotate by 90 degrees clockwise.
20499 cclock
20501 Rotate by 90 degrees counterclockwise.
20502 clock_flip
20504 Rotate by 90 degrees clockwise and vertically flip.
20505 passthrough
20507 Do not apply the transposition if the input geometry matches the
20508 one specified by the specified value. It accepts the following
20509 values:
20510 none
20512 Always apply transposition. (default)
20513 portrait
20515 Preserve portrait geometry (when height >= width).
20516 landscape
20518 Preserve landscape geometry (when width >= height).
20519 trim
20521 Trim the input so that the output contains one continuous subpart of
20522 the input.
20523 It accepts the following parameters:
20525 start
20527 Specify the time of the start of the kept section, i.e. the frame
20528 with the timestamp start will be the first frame in the output.
20529 end Specify the time of the first frame that will be dropped, i.e. the
20531 frame immediately preceding the one with the timestamp end will be
20532 the last frame in the output.
20533 start_pts
20535 This is the same as start, except this option sets the start
20536 timestamp in timebase units instead of seconds.
20537 end_pts
20539 This is the same as end, except this option sets the end timestamp
20540 in timebase units instead of seconds.
20541 duration
20543 The maximum duration of the output in seconds.
20544 start_frame
20546 The number of the first frame that should be passed to the output.
20547 end_frame
20549 The number of the first frame that should be dropped.
20550 start, end, and duration are expressed as time duration specifications;
20552 see the Time duration section in the ffmpeg-utils(1) manual for the
20553 accepted syntax.
20554 Note that the first two sets of the start/end options and the duration
20556 option look at the frame timestamp, while the _frame variants simply
20557 count the frames that pass through the filter. Also note that this
20558 filter does not modify the timestamps. If you wish for the output
20559 timestamps to start at zero, insert a setpts filter after the trim
20560 filter.
20561 If multiple start or end options are set, this filter tries to be
20563 greedy and keep all the frames that match at least one of the specified
20564 constraints. To keep only the part that matches all the constraints at
20565 once, chain multiple trim filters.
20566 The defaults are such that all the input is kept. So it is possible to
20568 set e.g. just the end values to keep everything before the specified
20569 time.
20570 Examples:
20572 • Drop everything except the second minute of input:
20574 ffmpeg -i INPUT -vf trim=60:120
20576 • Keep only the first second:
20578 ffmpeg -i INPUT -vf trim=duration=1
20580 unpremultiply
20582 Apply alpha unpremultiply effect to input video stream using first
20583 plane of second stream as alpha.
20584 Both streams must have same dimensions and same pixel format.
20586 The filter accepts the following option:
20588 planes
20590 Set which planes will be processed, unprocessed planes will be
20591 copied. By default value 0xf, all planes will be processed.
20592 If the format has 1 or 2 components, then luma is bit 0. If the
20594 format has 3 or 4 components: for RGB formats bit 0 is green, bit 1
20595 is blue and bit 2 is red; for YUV formats bit 0 is luma, bit 1 is
20596 chroma-U and bit 2 is chroma-V. If present, the alpha channel is
20597 always the last bit.
20598 inplace
20600 Do not require 2nd input for processing, instead use alpha plane
20601 from input stream.
20602 unsharp
20604 Sharpen or blur the input video.
20605 It accepts the following parameters:
20607 luma_msize_x, lx
20609 Set the luma matrix horizontal size. It must be an odd integer
20610 between 3 and 23. The default value is 5.
20611 luma_msize_y, ly
20613 Set the luma matrix vertical size. It must be an odd integer
20614 between 3 and 23. The default value is 5.
20615 luma_amount, la
20617 Set the luma effect strength. It must be a floating point number,
20618 reasonable values lay between -1.5 and 1.5.
20619 Negative values will blur the input video, while positive values
20621 will sharpen it, a value of zero will disable the effect.
20622 Default value is 1.0.
20624 chroma_msize_x, cx
20626 Set the chroma matrix horizontal size. It must be an odd integer
20627 between 3 and 23. The default value is 5.
20628 chroma_msize_y, cy
20630 Set the chroma matrix vertical size. It must be an odd integer
20631 between 3 and 23. The default value is 5.
20632 chroma_amount, ca
20634 Set the chroma effect strength. It must be a floating point number,
20635 reasonable values lay between -1.5 and 1.5.
20636 Negative values will blur the input video, while positive values
20638 will sharpen it, a value of zero will disable the effect.
20639 Default value is 0.0.
20641 alpha_msize_x, ax
20643 Set the alpha matrix horizontal size. It must be an odd integer
20644 between 3 and 23. The default value is 5.
20645 alpha_msize_y, ay
20647 Set the alpha matrix vertical size. It must be an odd integer
20648 between 3 and 23. The default value is 5.
20649 alpha_amount, aa
20651 Set the alpha effect strength. It must be a floating point number,
20652 reasonable values lay between -1.5 and 1.5.
20653 Negative values will blur the input video, while positive values
20655 will sharpen it, a value of zero will disable the effect.
20656 Default value is 0.0.
20658 All parameters are optional and default to the equivalent of the string
20660 '5:5:1.0:5:5:0.0'.
20661 Examples
20663 • Apply strong luma sharpen effect:
20665 unsharp=luma_msize_x=7:luma_msize_y=7:luma_amount=2.5
20667 • Apply a strong blur of both luma and chroma parameters:
20669 unsharp=7:7:-2:7:7:-2
20671 untile
20673 Decompose a video made of tiled images into the individual images.
20674 The frame rate of the output video is the frame rate of the input video
20676 multiplied by the number of tiles.
20677 This filter does the reverse of tile.
20679 The filter accepts the following options:
20681 layout
20683 Set the grid size (i.e. the number of lines and columns). For the
20684 syntax of this option, check the "Video size" section in the
20685 ffmpeg-utils manual.
20686 Examples
20688 • Produce a 1-second video from a still image file made of 25 frames
20690 stacked vertically, like an analogic film reel:
20691 ffmpeg -r 1 -i image.jpg -vf untile=1x25 movie.mkv
20693 uspp
20695 Apply ultra slow/simple postprocessing filter that compresses and
20696 decompresses the image at several (or - in the case of quality level 8
20697 - all) shifts and average the results.
20698 The way this differs from the behavior of spp is that uspp actually
20700 encodes & decodes each case with libavcodec Snow, whereas spp uses a
20701 simplified intra only 8x8 DCT similar to MJPEG.
20702 This filter is only available in ffmpeg version 4.4 or earlier.
20704 The filter accepts the following options:
20706 quality
20708 Set quality. This option defines the number of levels for
20709 averaging. It accepts an integer in the range 0-8. If set to 0, the
20710 filter will have no effect. A value of 8 means the higher quality.
20711 For each increment of that value the speed drops by a factor of
20712 approximately 2. Default value is 3.
20713 qp Force a constant quantization parameter. If not set, the filter
20715 will use the QP from the video stream (if available).
20716 v360
20718 Convert 360 videos between various formats.
20719 The filter accepts the following options:
20721 input
20723 output
20724 Set format of the input/output video.
20725 Available formats:
20727 e
20729 equirect
20730 Equirectangular projection.
20731 c3x2
20733 c6x1
20734 c1x6
20735 Cubemap with 3x2/6x1/1x6 layout.
20736 Format specific options:
20738 in_pad
20740 out_pad
20741 Set padding proportion for the input/output cubemap. Values
20742 in decimals.
20743 Example values:
20745 0 No padding.
20747 0.01
20749 1% of face is padding. For example, with 1920x1280
20750 resolution face size would be 640x640 and padding would
20751 be 3 pixels from each side. (640 * 0.01 = 6 pixels)
20752 Default value is @samp{0}. Maximum value is @samp{0.1}.
20754 fin_pad
20756 fout_pad
20757 Set fixed padding for the input/output cubemap. Values in
20758 pixels.
20759 Default value is @samp{0}. If greater than zero it
20761 overrides other padding options.
20762 in_forder
20764 out_forder
20765 Set order of faces for the input/output cubemap. Choose one
20766 direction for each position.
20767 Designation of directions:
20769 r right
20771 l left
20773 u up
20775 d down
20777 f forward
20779 b back
20781 Default value is @samp{rludfb}.
20783 in_frot
20785 out_frot
20786 Set rotation of faces for the input/output cubemap. Choose
20787 one angle for each position.
20788 Designation of angles:
20790 0 0 degrees clockwise
20792 1 90 degrees clockwise
20794 2 180 degrees clockwise
20796 3 270 degrees clockwise
20798 Default value is @samp{000000}.
20800 eac Equi-Angular Cubemap.
20802 flat
20804 gnomonic
20805 rectilinear
20806 Regular video.
20807 Format specific options:
20809 h_fov
20811 v_fov
20812 d_fov
20813 Set output horizontal/vertical/diagonal field of view.
20814 Values in degrees.
20815 If diagonal field of view is set it overrides horizontal
20817 and vertical field of view.
20818 ih_fov
20820 iv_fov
20821 id_fov
20822 Set input horizontal/vertical/diagonal field of view.
20823 Values in degrees.
20824 If diagonal field of view is set it overrides horizontal
20826 and vertical field of view.
20827 dfisheye
20829 Dual fisheye.
20830 Format specific options:
20832 h_fov
20834 v_fov
20835 d_fov
20836 Set output horizontal/vertical/diagonal field of view.
20837 Values in degrees.
20838 If diagonal field of view is set it overrides horizontal
20840 and vertical field of view.
20841 ih_fov
20843 iv_fov
20844 id_fov
20845 Set input horizontal/vertical/diagonal field of view.
20846 Values in degrees.
20847 If diagonal field of view is set it overrides horizontal
20849 and vertical field of view.
20850 barrel
20852 fb
20853 barrelsplit
20854 Facebook's 360 formats.
20855 sg Stereographic format.
20857 Format specific options:
20859 h_fov
20861 v_fov
20862 d_fov
20863 Set output horizontal/vertical/diagonal field of view.
20864 Values in degrees.
20865 If diagonal field of view is set it overrides horizontal
20867 and vertical field of view.
20868 ih_fov
20870 iv_fov
20871 id_fov
20872 Set input horizontal/vertical/diagonal field of view.
20873 Values in degrees.
20874 If diagonal field of view is set it overrides horizontal
20876 and vertical field of view.
20877 mercator
20879 Mercator format.
20880 ball
20882 Ball format, gives significant distortion toward the back.
20883 hammer
20885 Hammer-Aitoff map projection format.
20886 sinusoidal
20888 Sinusoidal map projection format.
20889 fisheye
20891 Fisheye projection.
20892 Format specific options:
20894 h_fov
20896 v_fov
20897 d_fov
20898 Set output horizontal/vertical/diagonal field of view.
20899 Values in degrees.
20900 If diagonal field of view is set it overrides horizontal
20902 and vertical field of view.
20903 ih_fov
20905 iv_fov
20906 id_fov
20907 Set input horizontal/vertical/diagonal field of view.
20908 Values in degrees.
20909 If diagonal field of view is set it overrides horizontal
20911 and vertical field of view.
20912 pannini
20914 Pannini projection.
20915 Format specific options:
20917 h_fov
20919 Set output pannini parameter.
20920 ih_fov
20922 Set input pannini parameter.
20923 cylindrical
20925 Cylindrical projection.
20926 Format specific options:
20928 h_fov
20930 v_fov
20931 d_fov
20932 Set output horizontal/vertical/diagonal field of view.
20933 Values in degrees.
20934 If diagonal field of view is set it overrides horizontal
20936 and vertical field of view.
20937 ih_fov
20939 iv_fov
20940 id_fov
20941 Set input horizontal/vertical/diagonal field of view.
20942 Values in degrees.
20943 If diagonal field of view is set it overrides horizontal
20945 and vertical field of view.
20946 perspective
20948 Perspective projection. (output only)
20949 Format specific options:
20951 v_fov
20953 Set perspective parameter.
20954 tetrahedron
20956 Tetrahedron projection.
20957 tsp Truncated square pyramid projection.
20959 he
20961 hequirect
20962 Half equirectangular projection.
20963 equisolid
20965 Equisolid format.
20966 Format specific options:
20968 h_fov
20970 v_fov
20971 d_fov
20972 Set output horizontal/vertical/diagonal field of view.
20973 Values in degrees.
20974 If diagonal field of view is set it overrides horizontal
20976 and vertical field of view.
20977 ih_fov
20979 iv_fov
20980 id_fov
20981 Set input horizontal/vertical/diagonal field of view.
20982 Values in degrees.
20983 If diagonal field of view is set it overrides horizontal
20985 and vertical field of view.
20986 og Orthographic format.
20988 Format specific options:
20990 h_fov
20992 v_fov
20993 d_fov
20994 Set output horizontal/vertical/diagonal field of view.
20995 Values in degrees.
20996 If diagonal field of view is set it overrides horizontal
20998 and vertical field of view.
20999 ih_fov
21001 iv_fov
21002 id_fov
21003 Set input horizontal/vertical/diagonal field of view.
21004 Values in degrees.
21005 If diagonal field of view is set it overrides horizontal
21007 and vertical field of view.
21008 octahedron
21010 Octahedron projection.
21011 cylindricalea
21013 Cylindrical Equal Area projection.
21014 interp
21016 Set interpolation method.Note: more complex interpolation methods
21017 require much more memory to run.
21018 Available methods:
21020 near
21022 nearest
21023 Nearest neighbour.
21024 line
21026 linear
21027 Bilinear interpolation.
21028 lagrange9
21030 Lagrange9 interpolation.
21031 cube
21033 cubic
21034 Bicubic interpolation.
21035 lanc
21037 lanczos
21038 Lanczos interpolation.
21039 sp16
21041 spline16
21042 Spline16 interpolation.
21043 gauss
21045 gaussian
21046 Gaussian interpolation.
21047 mitchell
21049 Mitchell interpolation.
21050 Default value is @samp{line}.
21052 w
21054 h Set the output video resolution.
21055 Default resolution depends on formats.
21057 in_stereo
21059 out_stereo
21060 Set the input/output stereo format.
21061 2d 2D mono
21063 sbs Side by side
21065 tb Top bottom
21067 Default value is @samp{2d} for input and output format.
21069 yaw
21071 pitch
21072 roll
21073 Set rotation for the output video. Values in degrees.
21074 rorder
21076 Set rotation order for the output video. Choose one item for each
21077 position.
21078 y, Y
21080 yaw
21081 p, P
21083 pitch
21084 r, R
21086 roll
21087 Default value is @samp{ypr}.
21089 h_flip
21091 v_flip
21092 d_flip
21093 Flip the output video horizontally(swaps
21094 left-right)/vertically(swaps up-down)/in-depth(swaps back-forward).
21095 Boolean values.
21096 ih_flip
21098 iv_flip
21099 Set if input video is flipped horizontally/vertically. Boolean
21100 values.
21101 in_trans
21103 Set if input video is transposed. Boolean value, by default
21104 disabled.
21105 out_trans
21107 Set if output video needs to be transposed. Boolean value, by
21108 default disabled.
21109 h_offset
21111 v_offset
21112 Set output horizontal/vertical off-axis offset. Default is set to
21113 0. Allowed range is from -1 to 1.
21114 alpha_mask
21116 Build mask in alpha plane for all unmapped pixels by marking them
21117 fully transparent. Boolean value, by default disabled.
21118 reset_rot
21120 Reset rotation of output video. Boolean value, by default disabled.
21121 Examples
21123 • Convert equirectangular video to cubemap with 3x2 layout and 1%
21125 padding using bicubic interpolation:
21126 ffmpeg -i input.mkv -vf v360=e:c3x2:cubic:out_pad=0.01 output.mkv
21128 • Extract back view of Equi-Angular Cubemap:
21130 ffmpeg -i input.mkv -vf v360=eac:flat:yaw=180 output.mkv
21132 • Convert transposed and horizontally flipped Equi-Angular Cubemap in
21134 side-by-side stereo format to equirectangular top-bottom stereo
21135 format:
21136 v360=eac:equirect:in_stereo=sbs:in_trans=1:ih_flip=1:out_stereo=tb
21138 Commands
21140 This filter supports subset of above options as commands.
21142 vaguedenoiser
21144 Apply a wavelet based denoiser.
21145 It transforms each frame from the video input into the wavelet domain,
21147 using Cohen-Daubechies-Feauveau 9/7. Then it applies some filtering to
21148 the obtained coefficients. It does an inverse wavelet transform after.
21149 Due to wavelet properties, it should give a nice smoothed result, and
21150 reduced noise, without blurring picture features.
21151 This filter accepts the following options:
21153 threshold
21155 The filtering strength. The higher, the more filtered the video
21156 will be. Hard thresholding can use a higher threshold than soft
21157 thresholding before the video looks overfiltered. Default value is
21158 2.
21159 method
21161 The filtering method the filter will use.
21162 It accepts the following values:
21164 hard
21166 All values under the threshold will be zeroed.
21167 soft
21169 All values under the threshold will be zeroed. All values above
21170 will be reduced by the threshold.
21171 garrote
21173 Scales or nullifies coefficients - intermediary between (more)
21174 soft and (less) hard thresholding.
21175 Default is garrote.
21177 nsteps
21179 Number of times, the wavelet will decompose the picture. Picture
21180 can't be decomposed beyond a particular point (typically, 8 for a
21181 640x480 frame - as 2^9 = 512 > 480). Valid values are integers
21182 between 1 and 32. Default value is 6.
21183 percent
21185 Partial of full denoising (limited coefficients shrinking), from 0
21186 to 100. Default value is 85.
21187 planes
21189 A list of the planes to process. By default all planes are
21190 processed.
21191 type
21193 The threshold type the filter will use.
21194 It accepts the following values:
21196 universal
21198 Threshold used is same for all decompositions.
21199 bayes
21201 Threshold used depends also on each decomposition coefficients.
21202 Default is universal.
21204 varblur
21206 Apply variable blur filter by using 2nd video stream to set blur
21207 radius. The 2nd stream must have the same dimensions.
21208 This filter accepts the following options:
21210 min_r
21212 Set min allowed radius. Allowed range is from 0 to 254. Default is
21213 0.
21214 max_r
21216 Set max allowed radius. Allowed range is from 1 to 255. Default is
21217 8.
21218 planes
21220 Set which planes to process. By default, all are used.
21221 The "varblur" filter also supports the framesync options.
21223 Commands
21225 This filter supports all the above options as commands.
21227 vectorscope
21229 Display 2 color component values in the two dimensional graph (which is
21230 called a vectorscope).
21231 This filter accepts the following options:
21233 mode, m
21235 Set vectorscope mode.
21236 It accepts the following values:
21238 gray
21240 tint
21241 Gray values are displayed on graph, higher brightness means
21242 more pixels have same component color value on location in
21243 graph. This is the default mode.
21244 color
21246 Gray values are displayed on graph. Surrounding pixels values
21247 which are not present in video frame are drawn in gradient of 2
21248 color components which are set by option "x" and "y". The 3rd
21249 color component is static.
21250 color2
21252 Actual color components values present in video frame are
21253 displayed on graph.
21254 color3
21256 Similar as color2 but higher frequency of same values "x" and
21257 "y" on graph increases value of another color component, which
21258 is luminance by default values of "x" and "y".
21259 color4
21261 Actual colors present in video frame are displayed on graph. If
21262 two different colors map to same position on graph then color
21263 with higher value of component not present in graph is picked.
21264 color5
21266 Gray values are displayed on graph. Similar to "color" but with
21267 3rd color component picked from radial gradient.
21268 x Set which color component will be represented on X-axis. Default is
21270 1.
21271 y Set which color component will be represented on Y-axis. Default is
21273 2.
21274 intensity, i
21276 Set intensity, used by modes: gray, color, color3 and color5 for
21277 increasing brightness of color component which represents frequency
21278 of (X, Y) location in graph.
21279 envelope, e
21281 none
21282 No envelope, this is default.
21283 instant
21285 Instant envelope, even darkest single pixel will be clearly
21286 highlighted.
21287 peak
21289 Hold maximum and minimum values presented in graph over time.
21290 This way you can still spot out of range values without
21291 constantly looking at vectorscope.
21292 peak+instant
21294 Peak and instant envelope combined together.
21295 graticule, g
21297 Set what kind of graticule to draw.
21298 none
21300 green
21301 color
21302 invert
21303 opacity, o
21304 Set graticule opacity.
21305 flags, f
21307 Set graticule flags.
21308 white
21310 Draw graticule for white point.
21311 black
21313 Draw graticule for black point.
21314 name
21316 Draw color points short names.
21317 bgopacity, b
21319 Set background opacity.
21320 lthreshold, l
21322 Set low threshold for color component not represented on X or Y
21323 axis. Values lower than this value will be ignored. Default is 0.
21324 Note this value is multiplied with actual max possible value one
21325 pixel component can have. So for 8-bit input and low threshold
21326 value of 0.1 actual threshold is 0.1 * 255 = 25.
21327 hthreshold, h
21329 Set high threshold for color component not represented on X or Y
21330 axis. Values higher than this value will be ignored. Default is 1.
21331 Note this value is multiplied with actual max possible value one
21332 pixel component can have. So for 8-bit input and high threshold
21333 value of 0.9 actual threshold is 0.9 * 255 = 230.
21334 colorspace, c
21336 Set what kind of colorspace to use when drawing graticule.
21337 auto
21339 601
21340 709
21341 Default is auto.
21343 tint0, t0
21345 tint1, t1
21346 Set color tint for gray/tint vectorscope mode. By default both
21347 options are zero. This means no tint, and output will remain gray.
21348 vidstabdetect
21350 Analyze video stabilization/deshaking. Perform pass 1 of 2, see
21351 vidstabtransform for pass 2.
21352 This filter generates a file with relative translation and rotation
21354 transform information about subsequent frames, which is then used by
21355 the vidstabtransform filter.
21356 To enable compilation of this filter you need to configure FFmpeg with
21358 "--enable-libvidstab".
21359 This filter accepts the following options:
21361 result
21363 Set the path to the file used to write the transforms information.
21364 Default value is transforms.trf.
21365 shakiness
21367 Set how shaky the video is and how quick the camera is. It accepts
21368 an integer in the range 1-10, a value of 1 means little shakiness,
21369 a value of 10 means strong shakiness. Default value is 5.
21370 accuracy
21372 Set the accuracy of the detection process. It must be a value in
21373 the range 1-15. A value of 1 means low accuracy, a value of 15
21374 means high accuracy. Default value is 15.
21375 stepsize
21377 Set stepsize of the search process. The region around minimum is
21378 scanned with 1 pixel resolution. Default value is 6.
21379 mincontrast
21381 Set minimum contrast. Below this value a local measurement field is
21382 discarded. Must be a floating point value in the range 0-1. Default
21383 value is 0.3.
21384 tripod
21386 Set reference frame number for tripod mode.
21387 If enabled, the motion of the frames is compared to a reference
21389 frame in the filtered stream, identified by the specified number.
21390 The idea is to compensate all movements in a more-or-less static
21391 scene and keep the camera view absolutely still.
21392 If set to 0, it is disabled. The frames are counted starting from
21394 1.
21395 show
21397 Show fields and transforms in the resulting frames. It accepts an
21398 integer in the range 0-2. Default value is 0, which disables any
21399 visualization.
21400 Examples
21402 • Use default values:
21404 vidstabdetect
21406 • Analyze strongly shaky movie and put the results in file
21408 mytransforms.trf:
21409 vidstabdetect=shakiness=10:accuracy=15:result="mytransforms.trf"
21411 • Visualize the result of internal transformations in the resulting
21413 video:
21414 vidstabdetect=show=1
21416 • Analyze a video with medium shakiness using ffmpeg:
21418 ffmpeg -i input -vf vidstabdetect=shakiness=5:show=1 dummy.avi
21420 vidstabtransform
21422 Video stabilization/deshaking: pass 2 of 2, see vidstabdetect for pass
21423 1.
21424 Read a file with transform information for each frame and
21426 apply/compensate them. Together with the vidstabdetect filter this can
21427 be used to deshake videos. See also
21428 <http://public.hronopik.de/vid.stab>. It is important to also use the
21429 unsharp filter, see below.
21430 To enable compilation of this filter you need to configure FFmpeg with
21432 "--enable-libvidstab".
21433 Options
21435 input
21437 Set path to the file used to read the transforms. Default value is
21438 transforms.trf.
21439 smoothing
21441 Set the number of frames (value*2 + 1) used for lowpass filtering
21442 the camera movements. Default value is 10.
21443 For example a number of 10 means that 21 frames are used (10 in the
21445 past and 10 in the future) to smoothen the motion in the video. A
21446 larger value leads to a smoother video, but limits the acceleration
21447 of the camera (pan/tilt movements). 0 is a special case where a
21448 static camera is simulated.
21449 optalgo
21451 Set the camera path optimization algorithm.
21452 Accepted values are:
21454 gauss
21456 gaussian kernel low-pass filter on camera motion (default)
21457 avg averaging on transformations
21459 maxshift
21461 Set maximal number of pixels to translate frames. Default value is
21462 -1, meaning no limit.
21463 maxangle
21465 Set maximal angle in radians (degree*PI/180) to rotate frames.
21466 Default value is -1, meaning no limit.
21467 crop
21469 Specify how to deal with borders that may be visible due to
21470 movement compensation.
21471 Available values are:
21473 keep
21475 keep image information from previous frame (default)
21476 black
21478 fill the border black
21479 invert
21481 Invert transforms if set to 1. Default value is 0.
21482 relative
21484 Consider transforms as relative to previous frame if set to 1,
21485 absolute if set to 0. Default value is 0.
21486 zoom
21488 Set percentage to zoom. A positive value will result in a zoom-in
21489 effect, a negative value in a zoom-out effect. Default value is 0
21490 (no zoom).
21491 optzoom
21493 Set optimal zooming to avoid borders.
21494 Accepted values are:
21496 0 disabled
21498 1 optimal static zoom value is determined (only very strong
21500 movements will lead to visible borders) (default)
21501 2 optimal adaptive zoom value is determined (no borders will be
21503 visible), see zoomspeed
21504 Note that the value given at zoom is added to the one calculated
21506 here.
21507 zoomspeed
21509 Set percent to zoom maximally each frame (enabled when optzoom is
21510 set to 2). Range is from 0 to 5, default value is 0.25.
21511 interpol
21513 Specify type of interpolation.
21514 Available values are:
21516 no no interpolation
21518 linear
21520 linear only horizontal
21521 bilinear
21523 linear in both directions (default)
21524 bicubic
21526 cubic in both directions (slow)
21527 tripod
21529 Enable virtual tripod mode if set to 1, which is equivalent to
21530 "relative=0:smoothing=0". Default value is 0.
21531 Use also "tripod" option of vidstabdetect.
21533 debug
21535 Increase log verbosity if set to 1. Also the detected global
21536 motions are written to the temporary file global_motions.trf.
21537 Default value is 0.
21538 Examples
21540 • Use ffmpeg for a typical stabilization with default values:
21542 ffmpeg -i inp.mpeg -vf vidstabtransform,unsharp=5:5:0.8:3:3:0.4 inp_stabilized.mpeg
21544 Note the use of the unsharp filter which is always recommended.
21546 • Zoom in a bit more and load transform data from a given file:
21548 vidstabtransform=zoom=5:input="mytransforms.trf"
21550 • Smoothen the video even more:
21552 vidstabtransform=smoothing=30
21554 vflip
21556 Flip the input video vertically.
21557 For example, to vertically flip a video with ffmpeg:
21559 ffmpeg -i in.avi -vf "vflip" out.avi
21561 vfrdet
21563 Detect variable frame rate video.
21564 This filter tries to detect if the input is variable or constant frame
21566 rate.
21567 At end it will output number of frames detected as having variable
21569 delta pts, and ones with constant delta pts. If there was frames with
21570 variable delta, than it will also show min, max and average delta
21571 encountered.
21572 vibrance
21574 Boost or alter saturation.
21575 The filter accepts the following options:
21577 intensity
21579 Set strength of boost if positive value or strength of alter if
21580 negative value. Default is 0. Allowed range is from -2 to 2.
21581 rbal
21583 Set the red balance. Default is 1. Allowed range is from -10 to 10.
21584 gbal
21586 Set the green balance. Default is 1. Allowed range is from -10 to
21587 10.
21588 bbal
21590 Set the blue balance. Default is 1. Allowed range is from -10 to
21591 10.
21592 rlum
21594 Set the red luma coefficient.
21595 glum
21597 Set the green luma coefficient.
21598 blum
21600 Set the blue luma coefficient.
21601 alternate
21603 If "intensity" is negative and this is set to 1, colors will
21604 change, otherwise colors will be less saturated, more towards gray.
21605 Commands
21607 This filter supports the all above options as commands.
21609 vif
21611 Obtain the average VIF (Visual Information Fidelity) between two input
21612 videos.
21613 This filter takes two input videos.
21615 Both input videos must have the same resolution and pixel format for
21617 this filter to work correctly. Also it assumes that both inputs have
21618 the same number of frames, which are compared one by one.
21619 The obtained average VIF score is printed through the logging system.
21621 The filter stores the calculated VIF score of each frame.
21623 This filter also supports the framesync options.
21625 In the below example the input file main.mpg being processed is
21627 compared with the reference file ref.mpg.
21628 ffmpeg -i main.mpg -i ref.mpg -lavfi vif -f null -
21630 vignette
21632 Make or reverse a natural vignetting effect.
21633 The filter accepts the following options:
21635 angle, a
21637 Set lens angle expression as a number of radians.
21638 The value is clipped in the "[0,PI/2]" range.
21640 Default value: "PI/5"
21642 x0
21644 y0 Set center coordinates expressions. Respectively "w/2" and "h/2" by
21645 default.
21646 mode
21648 Set forward/backward mode.
21649 Available modes are:
21651 forward
21653 The larger the distance from the central point, the darker the
21654 image becomes.
21655 backward
21657 The larger the distance from the central point, the brighter
21658 the image becomes. This can be used to reverse a vignette
21659 effect, though there is no automatic detection to extract the
21660 lens angle and other settings (yet). It can also be used to
21661 create a burning effect.
21662 Default value is forward.
21664 eval
21666 Set evaluation mode for the expressions (angle, x0, y0).
21667 It accepts the following values:
21669 init
21671 Evaluate expressions only once during the filter
21672 initialization.
21673 frame
21675 Evaluate expressions for each incoming frame. This is way
21676 slower than the init mode since it requires all the scalers to
21677 be re-computed, but it allows advanced dynamic expressions.
21678 Default value is init.
21680 dither
21682 Set dithering to reduce the circular banding effects. Default is 1
21683 (enabled).
21684 aspect
21686 Set vignette aspect. This setting allows one to adjust the shape of
21687 the vignette. Setting this value to the SAR of the input will make
21688 a rectangular vignetting following the dimensions of the video.
21689 Default is "1/1".
21691 Expressions
21693 The alpha, x0 and y0 expressions can contain the following parameters.
21695 w
21697 h input width and height
21698 n the number of input frame, starting from 0
21700 pts the PTS (Presentation TimeStamp) time of the filtered video frame,
21702 expressed in TB units, NAN if undefined
21703 r frame rate of the input video, NAN if the input frame rate is
21705 unknown
21706 t the PTS (Presentation TimeStamp) of the filtered video frame,
21708 expressed in seconds, NAN if undefined
21709 tb time base of the input video
21711 Examples
21713 • Apply simple strong vignetting effect:
21715 vignette=PI/4
21717 • Make a flickering vignetting:
21719 vignette='PI/4+random(1)*PI/50':eval=frame
21721 vmafmotion
21723 Obtain the average VMAF motion score of a video. It is one of the
21724 component metrics of VMAF.
21725 The obtained average motion score is printed through the logging
21727 system.
21728 The filter accepts the following options:
21730 stats_file
21732 If specified, the filter will use the named file to save the motion
21733 score of each frame with respect to the previous frame. When
21734 filename equals "-" the data is sent to standard output.
21735 Example:
21737 ffmpeg -i ref.mpg -vf vmafmotion -f null -
21739 vstack
21741 Stack input videos vertically.
21742 All streams must be of same pixel format and of same width.
21744 Note that this filter is faster than using overlay and pad filter to
21746 create same output.
21747 The filter accepts the following options:
21749 inputs
21751 Set number of input streams. Default is 2.
21752 shortest
21754 If set to 1, force the output to terminate when the shortest input
21755 terminates. Default value is 0.
21756 w3fdif
21758 Deinterlace the input video ("w3fdif" stands for "Weston 3 Field
21759 Deinterlacing Filter").
21760 Based on the process described by Martin Weston for BBC R&D, and
21762 implemented based on the de-interlace algorithm written by Jim
21763 Easterbrook for BBC R&D, the Weston 3 field deinterlacing filter uses
21764 filter coefficients calculated by BBC R&D.
21765 This filter uses field-dominance information in frame to decide which
21767 of each pair of fields to place first in the output. If it gets it
21768 wrong use setfield filter before "w3fdif" filter.
21769 There are two sets of filter coefficients, so called "simple" and
21771 "complex". Which set of filter coefficients is used can be set by
21772 passing an optional parameter:
21773 filter
21775 Set the interlacing filter coefficients. Accepts one of the
21776 following values:
21777 simple
21779 Simple filter coefficient set.
21780 complex
21782 More-complex filter coefficient set.
21783 Default value is complex.
21785 mode
21787 The interlacing mode to adopt. It accepts one of the following
21788 values:
21789 frame
21791 Output one frame for each frame.
21792 field
21794 Output one frame for each field.
21795 The default value is "field".
21797 parity
21799 The picture field parity assumed for the input interlaced video. It
21800 accepts one of the following values:
21801 tff Assume the top field is first.
21803 bff Assume the bottom field is first.
21805 auto
21807 Enable automatic detection of field parity.
21808 The default value is "auto". If the interlacing is unknown or the
21810 decoder does not export this information, top field first will be
21811 assumed.
21812 deint
21814 Specify which frames to deinterlace. Accepts one of the following
21815 values:
21816 all Deinterlace all frames,
21818 interlaced
21820 Only deinterlace frames marked as interlaced.
21821 Default value is all.
21823 Commands
21825 This filter supports same commands as options.
21827 waveform
21829 Video waveform monitor.
21830 The waveform monitor plots color component intensity. By default
21832 luminance only. Each column of the waveform corresponds to a column of
21833 pixels in the source video.
21834 It accepts the following options:
21836 mode, m
21838 Can be either "row", or "column". Default is "column". In row
21839 mode, the graph on the left side represents color component value 0
21840 and the right side represents value = 255. In column mode, the top
21841 side represents color component value = 0 and bottom side
21842 represents value = 255.
21843 intensity, i
21845 Set intensity. Smaller values are useful to find out how many
21846 values of the same luminance are distributed across input
21847 rows/columns. Default value is 0.04. Allowed range is [0, 1].
21848 mirror, r
21850 Set mirroring mode. 0 means unmirrored, 1 means mirrored. In
21851 mirrored mode, higher values will be represented on the left side
21852 for "row" mode and at the top for "column" mode. Default is 1
21853 (mirrored).
21854 display, d
21856 Set display mode. It accepts the following values:
21857 overlay
21859 Presents information identical to that in the "parade", except
21860 that the graphs representing color components are superimposed
21861 directly over one another.
21862 This display mode makes it easier to spot relative differences
21864 or similarities in overlapping areas of the color components
21865 that are supposed to be identical, such as neutral whites,
21866 grays, or blacks.
21867 stack
21869 Display separate graph for the color components side by side in
21870 "row" mode or one below the other in "column" mode.
21871 parade
21873 Display separate graph for the color components side by side in
21874 "column" mode or one below the other in "row" mode.
21875 Using this display mode makes it easy to spot color casts in
21877 the highlights and shadows of an image, by comparing the
21878 contours of the top and the bottom graphs of each waveform.
21879 Since whites, grays, and blacks are characterized by exactly
21880 equal amounts of red, green, and blue, neutral areas of the
21881 picture should display three waveforms of roughly equal
21882 width/height. If not, the correction is easy to perform by
21883 making level adjustments the three waveforms.
21884 Default is "stack".
21886 components, c
21888 Set which color components to display. Default is 1, which means
21889 only luminance or red color component if input is in RGB
21890 colorspace. If is set for example to 7 it will display all 3 (if)
21891 available color components.
21892 envelope, e
21894 none
21895 No envelope, this is default.
21896 instant
21898 Instant envelope, minimum and maximum values presented in graph
21899 will be easily visible even with small "step" value.
21900 peak
21902 Hold minimum and maximum values presented in graph across time.
21903 This way you can still spot out of range values without
21904 constantly looking at waveforms.
21905 peak+instant
21907 Peak and instant envelope combined together.
21908 filter, f
21910 lowpass
21911 No filtering, this is default.
21912 flat
21914 Luma and chroma combined together.
21915 aflat
21917 Similar as above, but shows difference between blue and red
21918 chroma.
21919 xflat
21921 Similar as above, but use different colors.
21922 yflat
21924 Similar as above, but again with different colors.
21925 chroma
21927 Displays only chroma.
21928 color
21930 Displays actual color value on waveform.
21931 acolor
21933 Similar as above, but with luma showing frequency of chroma
21934 values.
21935 graticule, g
21937 Set which graticule to display.
21938 none
21940 Do not display graticule.
21941 green
21943 Display green graticule showing legal broadcast ranges.
21944 orange
21946 Display orange graticule showing legal broadcast ranges.
21947 invert
21949 Display invert graticule showing legal broadcast ranges.
21950 opacity, o
21952 Set graticule opacity.
21953 flags, fl
21955 Set graticule flags.
21956 numbers
21958 Draw numbers above lines. By default enabled.
21959 dots
21961 Draw dots instead of lines.
21962 scale, s
21964 Set scale used for displaying graticule.
21965 digital
21967 millivolts
21968 ire
21969 Default is digital.
21971 bgopacity, b
21973 Set background opacity.
21974 tint0, t0
21976 tint1, t1
21977 Set tint for output. Only used with lowpass filter and when
21978 display is not overlay and input pixel formats are not RGB.
21979 fitmode, fm
21981 Set sample aspect ratio of video output frames. Can be used to
21982 configure waveform so it is not streched too much in one of
21983 directions.
21984 none
21986 Set sample aspect ration to 1/1.
21987 size
21989 Set sample aspect ratio to match input size of video
21990 Default is none.
21992 weave, doubleweave
21994 The "weave" takes a field-based video input and join each two
21995 sequential fields into single frame, producing a new double height clip
21996 with half the frame rate and half the frame count.
21997 The "doubleweave" works same as "weave" but without halving frame rate
21999 and frame count.
22000 It accepts the following option:
22002 first_field
22004 Set first field. Available values are:
22005 top, t
22007 Set the frame as top-field-first.
22008 bottom, b
22010 Set the frame as bottom-field-first.
22011 Examples
22013 • Interlace video using select and separatefields filter:
22015 separatefields,select=eq(mod(n,4),0)+eq(mod(n,4),3),weave
22017 xbr
22019 Apply the xBR high-quality magnification filter which is designed for
22020 pixel art. It follows a set of edge-detection rules, see
22021 <https://forums.libretro.com/t/xbr-algorithm-tutorial/123>.
22022 It accepts the following option:
22024 n Set the scaling dimension: 2 for "2xBR", 3 for "3xBR" and 4 for
22026 "4xBR". Default is 3.
22027 xcorrelate
22029 Apply normalized cross-correlation between first and second input video
22030 stream.
22031 Second input video stream dimensions must be lower than first input
22033 video stream.
22034 The filter accepts the following options:
22036 planes
22038 Set which planes to process.
22039 secondary
22041 Set which secondary video frames will be processed from second
22042 input video stream, can be first or all. Default is all.
22043 The "xcorrelate" filter also supports the framesync options.
22045 xfade
22047 Apply cross fade from one input video stream to another input video
22048 stream. The cross fade is applied for specified duration.
22049 Both inputs must be constant frame-rate and have the same resolution,
22051 pixel format, frame rate and timebase.
22052 The filter accepts the following options:
22054 transition
22056 Set one of available transition effects:
22057 custom
22059 fade
22060 wipeleft
22061 wiperight
22062 wipeup
22063 wipedown
22064 slideleft
22065 slideright
22066 slideup
22067 slidedown
22068 circlecrop
22069 rectcrop
22070 distance
22071 fadeblack
22072 fadewhite
22073 radial
22074 smoothleft
22075 smoothright
22076 smoothup
22077 smoothdown
22078 circleopen
22079 circleclose
22080 vertopen
22081 vertclose
22082 horzopen
22083 horzclose
22084 dissolve
22085 pixelize
22086 diagtl
22087 diagtr
22088 diagbl
22089 diagbr
22090 hlslice
22091 hrslice
22092 vuslice
22093 vdslice
22094 hblur
22095 fadegrays
22096 wipetl
22097 wipetr
22098 wipebl
22099 wipebr
22100 squeezeh
22101 squeezev
22102 zoomin
22103 fadefast
22104 fadeslow
22105 Default transition effect is fade.
22107 duration
22109 Set cross fade duration in seconds. Range is 0 to 60 seconds.
22110 Default duration is 1 second.
22111 offset
22113 Set cross fade start relative to first input stream in seconds.
22114 Default offset is 0.
22115 expr
22117 Set expression for custom transition effect.
22118 The expressions can use the following variables and functions:
22120 X
22122 Y The coordinates of the current sample.
22123 W
22125 H The width and height of the image.
22126 P Progress of transition effect.
22128 PLANE
22130 Currently processed plane.
22131 A Return value of first input at current location and plane.
22133 B Return value of second input at current location and plane.
22135 a0(x, y)
22137 a1(x, y)
22138 a2(x, y)
22139 a3(x, y)
22140 Return the value of the pixel at location (x,y) of the
22141 first/second/third/fourth component of first input.
22142 b0(x, y)
22144 b1(x, y)
22145 b2(x, y)
22146 b3(x, y)
22147 Return the value of the pixel at location (x,y) of the
22148 first/second/third/fourth component of second input.
22149 Examples
22151 • Cross fade from one input video to another input video, with fade
22153 transition and duration of transition of 2 seconds starting at
22154 offset of 5 seconds:
22155 ffmpeg -i first.mp4 -i second.mp4 -filter_complex xfade=transition=fade:duration=2:offset=5 output.mp4
22157 xmedian
22159 Pick median pixels from several input videos.
22160 The filter accepts the following options:
22162 inputs
22164 Set number of inputs. Default is 3. Allowed range is from 3 to
22165 255. If number of inputs is even number, than result will be mean
22166 value between two median values.
22167 planes
22169 Set which planes to filter. Default value is 15, by which all
22170 planes are processed.
22171 percentile
22173 Set median percentile. Default value is 0.5. Default value of 0.5
22174 will pick always median values, while 0 will pick minimum values,
22175 and 1 maximum values.
22176 Commands
22178 This filter supports all above options as commands, excluding option
22180 "inputs".
22181 xstack
22183 Stack video inputs into custom layout.
22184 All streams must be of same pixel format.
22186 The filter accepts the following options:
22188 inputs
22190 Set number of input streams. Default is 2.
22191 layout
22193 Specify layout of inputs. This option requires the desired layout
22194 configuration to be explicitly set by the user. This sets position
22195 of each video input in output. Each input is separated by '|'. The
22196 first number represents the column, and the second number
22197 represents the row. Numbers start at 0 and are separated by '_'.
22198 Optionally one can use wX and hX, where X is video input from which
22199 to take width or height. Multiple values can be used when
22200 separated by '+'. In such case values are summed together.
22201 Note that if inputs are of different sizes gaps may appear, as not
22203 all of the output video frame will be filled. Similarly, videos can
22204 overlap each other if their position doesn't leave enough space for
22205 the full frame of adjoining videos.
22206 For 2 inputs, a default layout of "0_0|w0_0" (equivalent to
22208 "grid=2x1") is set. In all other cases, a layout or a grid must be
22209 set by the user. Either "grid" or "layout" can be specified at a
22210 time. Specifying both will result in an error.
22211 grid
22213 Specify a fixed size grid of inputs. This option is used to create
22214 a fixed size grid of the input streams. Set the grid size in the
22215 form "COLUMNSxROWS". There must be "ROWS * COLUMNS" input streams
22216 and they will be arranged as a grid with "ROWS" rows and "COLUMNS"
22217 columns. When using this option, each input stream within a row
22218 must have the same height and all the rows must have the same
22219 width.
22220 If "grid" is set, then "inputs" option is ignored and is implicitly
22222 set to "ROWS * COLUMNS".
22223 For 2 inputs, a default grid of "2x1" (equivalent to
22225 "layout=0_0|w0_0") is set. In all other cases, a layout or a grid
22226 must be set by the user. Either "grid" or "layout" can be specified
22227 at a time. Specifying both will result in an error.
22228 shortest
22230 If set to 1, force the output to terminate when the shortest input
22231 terminates. Default value is 0.
22232 fill
22234 If set to valid color, all unused pixels will be filled with that
22235 color. By default fill is set to none, so it is disabled.
22236 Examples
22238 • Display 4 inputs into 2x2 grid.
22240 Layout:
22242 input1(0, 0) | input3(w0, 0)
22244 input2(0, h0) | input4(w0, h0)
22245 xstack=inputs=4:layout=0_0|0_h0|w0_0|w0_h0
22249 Note that if inputs are of different sizes, gaps or overlaps may
22251 occur.
22252 • Display 4 inputs into 1x4 grid.
22254 Layout:
22256 input1(0, 0)
22258 input2(0, h0)
22259 input3(0, h0+h1)
22260 input4(0, h0+h1+h2)
22261 xstack=inputs=4:layout=0_0|0_h0|0_h0+h1|0_h0+h1+h2
22265 Note that if inputs are of different widths, unused space will
22267 appear.
22268 • Display 9 inputs into 3x3 grid.
22270 Layout:
22272 input1(0, 0) | input4(w0, 0) | input7(w0+w3, 0)
22274 input2(0, h0) | input5(w0, h0) | input8(w0+w3, h0)
22275 input3(0, h0+h1) | input6(w0, h0+h1) | input9(w0+w3, h0+h1)
22276 xstack=inputs=9:layout=0_0|0_h0|0_h0+h1|w0_0|w0_h0|w0_h0+h1|w0+w3_0|w0+w3_h0|w0+w3_h0+h1
22280 Note that if inputs are of different sizes, gaps or overlaps may
22282 occur.
22283 • Display 16 inputs into 4x4 grid.
22285 Layout:
22287 input1(0, 0) | input5(w0, 0) | input9 (w0+w4, 0) | input13(w0+w4+w8, 0)
22289 input2(0, h0) | input6(w0, h0) | input10(w0+w4, h0) | input14(w0+w4+w8, h0)
22290 input3(0, h0+h1) | input7(w0, h0+h1) | input11(w0+w4, h0+h1) | input15(w0+w4+w8, h0+h1)
22291 input4(0, h0+h1+h2)| input8(w0, h0+h1+h2)| input12(w0+w4, h0+h1+h2)| input16(w0+w4+w8, h0+h1+h2)
22292 xstack=inputs=16:layout=0_0|0_h0|0_h0+h1|0_h0+h1+h2|w0_0|w0_h0|w0_h0+h1|w0_h0+h1+h2|w0+w4_0|
22296 w0+w4_h0|w0+w4_h0+h1|w0+w4_h0+h1+h2|w0+w4+w8_0|w0+w4+w8_h0|w0+w4+w8_h0+h1|w0+w4+w8_h0+h1+h2
22297 Note that if inputs are of different sizes, gaps or overlaps may
22299 occur.
22300 yadif
22302 Deinterlace the input video ("yadif" means "yet another deinterlacing
22303 filter").
22304 It accepts the following parameters:
22306 mode
22308 The interlacing mode to adopt. It accepts one of the following
22309 values:
22310 0, send_frame
22312 Output one frame for each frame.
22313 1, send_field
22315 Output one frame for each field.
22316 2, send_frame_nospatial
22318 Like "send_frame", but it skips the spatial interlacing check.
22319 3, send_field_nospatial
22321 Like "send_field", but it skips the spatial interlacing check.
22322 The default value is "send_frame".
22324 parity
22326 The picture field parity assumed for the input interlaced video. It
22327 accepts one of the following values:
22328 0, tff
22330 Assume the top field is first.
22331 1, bff
22333 Assume the bottom field is first.
22334 -1, auto
22336 Enable automatic detection of field parity.
22337 The default value is "auto". If the interlacing is unknown or the
22339 decoder does not export this information, top field first will be
22340 assumed.
22341 deint
22343 Specify which frames to deinterlace. Accepts one of the following
22344 values:
22345 0, all
22347 Deinterlace all frames.
22348 1, interlaced
22350 Only deinterlace frames marked as interlaced.
22351 The default value is "all".
22353 yadif_cuda
22355 Deinterlace the input video using the yadif algorithm, but implemented
22356 in CUDA so that it can work as part of a GPU accelerated pipeline with
22357 nvdec and/or nvenc.
22358 It accepts the following parameters:
22360 mode
22362 The interlacing mode to adopt. It accepts one of the following
22363 values:
22364 0, send_frame
22366 Output one frame for each frame.
22367 1, send_field
22369 Output one frame for each field.
22370 2, send_frame_nospatial
22372 Like "send_frame", but it skips the spatial interlacing check.
22373 3, send_field_nospatial
22375 Like "send_field", but it skips the spatial interlacing check.
22376 The default value is "send_frame".
22378 parity
22380 The picture field parity assumed for the input interlaced video. It
22381 accepts one of the following values:
22382 0, tff
22384 Assume the top field is first.
22385 1, bff
22387 Assume the bottom field is first.
22388 -1, auto
22390 Enable automatic detection of field parity.
22391 The default value is "auto". If the interlacing is unknown or the
22393 decoder does not export this information, top field first will be
22394 assumed.
22395 deint
22397 Specify which frames to deinterlace. Accepts one of the following
22398 values:
22399 0, all
22401 Deinterlace all frames.
22402 1, interlaced
22404 Only deinterlace frames marked as interlaced.
22405 The default value is "all".
22407 yaepblur
22409 Apply blur filter while preserving edges ("yaepblur" means "yet another
22410 edge preserving blur filter"). The algorithm is described in "J. S.
22411 Lee, Digital image enhancement and noise filtering by use of local
22412 statistics, IEEE Trans. Pattern Anal. Mach. Intell. PAMI-2, 1980."
22413 It accepts the following parameters:
22415 radius, r
22417 Set the window radius. Default value is 3.
22418 planes, p
22420 Set which planes to filter. Default is only the first plane.
22421 sigma, s
22423 Set blur strength. Default value is 128.
22424 Commands
22426 This filter supports same commands as options.
22428 zoompan
22430 Apply Zoom & Pan effect.
22431 This filter accepts the following options:
22433 zoom, z
22435 Set the zoom expression. Range is 1-10. Default is 1.
22436 x
22438 y Set the x and y expression. Default is 0.
22439 d Set the duration expression in number of frames. This sets for how
22441 many number of frames effect will last for single input image.
22442 Default is 90.
22443 s Set the output image size, default is 'hd720'.
22445 fps Set the output frame rate, default is '25'.
22447 Each expression can contain the following constants:
22449 in_w, iw
22451 Input width.
22452 in_h, ih
22454 Input height.
22455 out_w, ow
22457 Output width.
22458 out_h, oh
22460 Output height.
22461 in Input frame count.
22463 on Output frame count.
22465 in_time, it
22467 The input timestamp expressed in seconds. It's NAN if the input
22468 timestamp is unknown.
22469 out_time, time, ot
22471 The output timestamp expressed in seconds.
22472 x
22474 y Last calculated 'x' and 'y' position from 'x' and 'y' expression
22475 for current input frame.
22476 px
22478 py 'x' and 'y' of last output frame of previous input frame or 0 when
22479 there was not yet such frame (first input frame).
22480 zoom
22482 Last calculated zoom from 'z' expression for current input frame.
22483 pzoom
22485 Last calculated zoom of last output frame of previous input frame.
22486 duration
22488 Number of output frames for current input frame. Calculated from
22489 'd' expression for each input frame.
22490 pduration
22492 number of output frames created for previous input frame
22493 a Rational number: input width / input height
22495 sar sample aspect ratio
22497 dar display aspect ratio
22499 Examples
22501 • Zoom in up to 1.5x and pan at same time to some spot near center of
22503 picture:
22504 zoompan=z='min(zoom+0.0015,1.5)':d=700:x='if(gte(zoom,1.5),x,x+1/a)':y='if(gte(zoom,1.5),y,y+1)':s=640x360
22506 • Zoom in up to 1.5x and pan always at center of picture:
22508 zoompan=z='min(zoom+0.0015,1.5)':d=700:x='iw/2-(iw/zoom/2)':y='ih/2-(ih/zoom/2)'
22510 • Same as above but without pausing:
22512 zoompan=z='min(max(zoom,pzoom)+0.0015,1.5)':d=1:x='iw/2-(iw/zoom/2)':y='ih/2-(ih/zoom/2)'
22514 • Zoom in 2x into center of picture only for the first second of the
22516 input video:
22517 zoompan=z='if(between(in_time,0,1),2,1)':d=1:x='iw/2-(iw/zoom/2)':y='ih/2-(ih/zoom/2)'
22519 zscale
22521 Scale (resize) the input video, using the z.lib library:
22522 <https://github.com/sekrit-twc/zimg>. To enable compilation of this
22523 filter, you need to configure FFmpeg with "--enable-libzimg".
22524 The zscale filter forces the output display aspect ratio to be the same
22526 as the input, by changing the output sample aspect ratio.
22527 If the input image format is different from the format requested by the
22529 next filter, the zscale filter will convert the input to the requested
22530 format.
22531 Options
22533 The filter accepts the following options.
22535 width, w
22537 height, h
22538 Set the output video dimension expression. Default value is the
22539 input dimension.
22540 If the width or w value is 0, the input width is used for the
22542 output. If the height or h value is 0, the input height is used for
22543 the output.
22544 If one and only one of the values is -n with n >= 1, the zscale
22546 filter will use a value that maintains the aspect ratio of the
22547 input image, calculated from the other specified dimension. After
22548 that it will, however, make sure that the calculated dimension is
22549 divisible by n and adjust the value if necessary.
22550 If both values are -n with n >= 1, the behavior will be identical
22552 to both values being set to 0 as previously detailed.
22553 See below for the list of accepted constants for use in the
22555 dimension expression.
22556 size, s
22558 Set the video size. For the syntax of this option, check the "Video
22559 size" section in the ffmpeg-utils manual.
22560 dither, d
22562 Set the dither type.
22563 Possible values are:
22565 none
22567 ordered
22568 random
22569 error_diffusion
22570 Default is none.
22572 filter, f
22574 Set the resize filter type.
22575 Possible values are:
22577 point
22579 bilinear
22580 bicubic
22581 spline16
22582 spline36
22583 lanczos
22584 Default is bilinear.
22586 range, r
22588 Set the color range.
22589 Possible values are:
22591 input
22593 limited
22594 full
22595 Default is same as input.
22597 primaries, p
22599 Set the color primaries.
22600 Possible values are:
22602 input
22604 709
22605 unspecified
22606 170m
22607 240m
22608 2020
22609 Default is same as input.
22611 transfer, t
22613 Set the transfer characteristics.
22614 Possible values are:
22616 input
22618 709
22619 unspecified
22620 601
22621 linear
22622 2020_10
22623 2020_12
22624 smpte2084
22625 iec61966-2-1
22626 arib-std-b67
22627 Default is same as input.
22629 matrix, m
22631 Set the colorspace matrix.
22632 Possible value are:
22634 input
22636 709
22637 unspecified
22638 470bg
22639 170m
22640 2020_ncl
22641 2020_cl
22642 Default is same as input.
22644 rangein, rin
22646 Set the input color range.
22647 Possible values are:
22649 input
22651 limited
22652 full
22653 Default is same as input.
22655 primariesin, pin
22657 Set the input color primaries.
22658 Possible values are:
22660 input
22662 709
22663 unspecified
22664 170m
22665 240m
22666 2020
22667 Default is same as input.
22669 transferin, tin
22671 Set the input transfer characteristics.
22672 Possible values are:
22674 input
22676 709
22677 unspecified
22678 601
22679 linear
22680 2020_10
22681 2020_12
22682 Default is same as input.
22684 matrixin, min
22686 Set the input colorspace matrix.
22687 Possible value are:
22689 input
22691 709
22692 unspecified
22693 470bg
22694 170m
22695 2020_ncl
22696 2020_cl
22697 chromal, c
22698 Set the output chroma location.
22699 Possible values are:
22701 input
22703 left
22704 center
22705 topleft
22706 top
22707 bottomleft
22708 bottom
22709 chromalin, cin
22710 Set the input chroma location.
22711 Possible values are:
22713 input
22715 left
22716 center
22717 topleft
22718 top
22719 bottomleft
22720 bottom
22721 npl Set the nominal peak luminance.
22722 param_a
22724 Parameter A for scaling filters. Parameter "b" for bicubic, and the
22725 number of filter taps for lanczos.
22726 param_b
22728 Parameter B for scaling filters. Parameter "c" for bicubic.
22729 The values of the w and h options are expressions containing the
22731 following constants:
22732 in_w
22734 in_h
22735 The input width and height
22736 iw
22738 ih These are the same as in_w and in_h.
22739 out_w
22741 out_h
22742 The output (scaled) width and height
22743 ow
22745 oh These are the same as out_w and out_h
22746 a The same as iw / ih
22748 sar input sample aspect ratio
22750 dar The input display aspect ratio. Calculated from "(iw / ih) * sar".
22752 hsub
22754 vsub
22755 horizontal and vertical input chroma subsample values. For example
22756 for the pixel format "yuv422p" hsub is 2 and vsub is 1.
22757 ohsub
22759 ovsub
22760 horizontal and vertical output chroma subsample values. For example
22761 for the pixel format "yuv422p" hsub is 2 and vsub is 1.
22762 Commands
22764 This filter supports the following commands:
22766 width, w
22768 height, h
22769 Set the output video dimension expression. The command accepts the
22770 same syntax of the corresponding option.
22771 If the specified expression is not valid, it is kept at its current
22773 value.
22774
22776 Below is a description of the currently available OpenCL video filters.
22777 To enable compilation of these filters you need to configure FFmpeg
22779 with "--enable-opencl".
22780 Running OpenCL filters requires you to initialize a hardware device and
22782 to pass that device to all filters in any filter graph.
22783 -init_hw_device opencl[=name][:device[,key=value...]]
22785 Initialise a new hardware device of type opencl called name, using
22786 the given device parameters.
22787 -filter_hw_device name
22789 Pass the hardware device called name to all filters in any filter
22790 graph.
22791 For more detailed information see
22793 <https://www.ffmpeg.org/ffmpeg.html#Advanced-Video-options>
22794 • Example of choosing the first device on the second platform and
22796 running avgblur_opencl filter with default parameters on it.
22797 -init_hw_device opencl=gpu:1.0 -filter_hw_device gpu -i INPUT -vf "hwupload, avgblur_opencl, hwdownload" OUTPUT
22799 Since OpenCL filters are not able to access frame data in normal
22801 memory, all frame data needs to be uploaded(hwupload) to hardware
22802 surfaces connected to the appropriate device before being used and then
22803 downloaded(hwdownload) back to normal memory. Note that hwupload will
22804 upload to a surface with the same layout as the software frame, so it
22805 may be necessary to add a format filter immediately before to get the
22806 input into the right format and hwdownload does not support all formats
22807 on the output - it may be necessary to insert an additional format
22808 filter immediately following in the graph to get the output in a
22809 supported format.
22810 avgblur_opencl
22812 Apply average blur filter.
22813 The filter accepts the following options:
22815 sizeX
22817 Set horizontal radius size. Range is "[1, 1024]" and default value
22818 is 1.
22819 planes
22821 Set which planes to filter. Default value is 0xf, by which all
22822 planes are processed.
22823 sizeY
22825 Set vertical radius size. Range is "[1, 1024]" and default value is
22826 0. If zero, "sizeX" value will be used.
22827 Example
22829 • Apply average blur filter with horizontal and vertical size of 3,
22831 setting each pixel of the output to the average value of the 7x7
22832 region centered on it in the input. For pixels on the edges of the
22833 image, the region does not extend beyond the image boundaries, and
22834 so out-of-range coordinates are not used in the calculations.
22835 -i INPUT -vf "hwupload, avgblur_opencl=3, hwdownload" OUTPUT
22837 boxblur_opencl
22839 Apply a boxblur algorithm to the input video.
22840 It accepts the following parameters:
22842 luma_radius, lr
22844 luma_power, lp
22845 chroma_radius, cr
22846 chroma_power, cp
22847 alpha_radius, ar
22848 alpha_power, ap
22849 A description of the accepted options follows.
22851 luma_radius, lr
22853 chroma_radius, cr
22854 alpha_radius, ar
22855 Set an expression for the box radius in pixels used for blurring
22856 the corresponding input plane.
22857 The radius value must be a non-negative number, and must not be
22859 greater than the value of the expression "min(w,h)/2" for the luma
22860 and alpha planes, and of "min(cw,ch)/2" for the chroma planes.
22861 Default value for luma_radius is "2". If not specified,
22863 chroma_radius and alpha_radius default to the corresponding value
22864 set for luma_radius.
22865 The expressions can contain the following constants:
22867 w
22869 h The input width and height in pixels.
22870 cw
22872 ch The input chroma image width and height in pixels.
22873 hsub
22875 vsub
22876 The horizontal and vertical chroma subsample values. For
22877 example, for the pixel format "yuv422p", hsub is 2 and vsub is
22878 1.
22879 luma_power, lp
22881 chroma_power, cp
22882 alpha_power, ap
22883 Specify how many times the boxblur filter is applied to the
22884 corresponding plane.
22885 Default value for luma_power is 2. If not specified, chroma_power
22887 and alpha_power default to the corresponding value set for
22888 luma_power.
22889 A value of 0 will disable the effect.
22891 Examples
22893 Apply boxblur filter, setting each pixel of the output to the average
22895 value of box-radiuses luma_radius, chroma_radius, alpha_radius for each
22896 plane respectively. The filter will apply luma_power, chroma_power,
22897 alpha_power times onto the corresponding plane. For pixels on the edges
22898 of the image, the radius does not extend beyond the image boundaries,
22899 and so out-of-range coordinates are not used in the calculations.
22900 • Apply a boxblur filter with the luma, chroma, and alpha radius set
22902 to 2 and luma, chroma, and alpha power set to 3. The filter will
22903 run 3 times with box-radius set to 2 for every plane of the image.
22904 -i INPUT -vf "hwupload, boxblur_opencl=luma_radius=2:luma_power=3, hwdownload" OUTPUT
22906 -i INPUT -vf "hwupload, boxblur_opencl=2:3, hwdownload" OUTPUT
22907 • Apply a boxblur filter with luma radius set to 2, luma_power to 1,
22909 chroma_radius to 4, chroma_power to 5, alpha_radius to 3 and
22910 alpha_power to 7.
22911 For the luma plane, a 2x2 box radius will be run once.
22913 For the chroma plane, a 4x4 box radius will be run 5 times.
22915 For the alpha plane, a 3x3 box radius will be run 7 times.
22917 -i INPUT -vf "hwupload, boxblur_opencl=2:1:4:5:3:7, hwdownload" OUTPUT
22919 colorkey_opencl
22921 RGB colorspace color keying.
22922 The filter accepts the following options:
22924 color
22926 The color which will be replaced with transparency.
22927 similarity
22929 Similarity percentage with the key color.
22930 0.01 matches only the exact key color, while 1.0 matches
22932 everything.
22933 blend
22935 Blend percentage.
22936 0.0 makes pixels either fully transparent, or not transparent at
22938 all.
22939 Higher values result in semi-transparent pixels, with a higher
22941 transparency the more similar the pixels color is to the key color.
22942 Examples
22944 • Make every semi-green pixel in the input transparent with some
22946 slight blending:
22947 -i INPUT -vf "hwupload, colorkey_opencl=green:0.3:0.1, hwdownload" OUTPUT
22949 convolution_opencl
22951 Apply convolution of 3x3, 5x5, 7x7 matrix.
22952 The filter accepts the following options:
22954 0m
22956 1m
22957 2m
22958 3m Set matrix for each plane. Matrix is sequence of 9, 25 or 49
22959 signed numbers. Default value for each plane is "0 0 0 0 1 0 0 0
22960 0".
22961 0rdiv
22963 1rdiv
22964 2rdiv
22965 3rdiv
22966 Set multiplier for calculated value for each plane. If unset or 0,
22967 it will be sum of all matrix elements. The option value must be a
22968 float number greater or equal to 0.0. Default value is 1.0.
22969 0bias
22971 1bias
22972 2bias
22973 3bias
22974 Set bias for each plane. This value is added to the result of the
22975 multiplication. Useful for making the overall image brighter or
22976 darker. The option value must be a float number greater or equal
22977 to 0.0. Default value is 0.0.
22978 Examples
22980 • Apply sharpen:
22982 -i INPUT -vf "hwupload, convolution_opencl=0 -1 0 -1 5 -1 0 -1 0:0 -1 0 -1 5 -1 0 -1 0:0 -1 0 -1 5 -1 0 -1 0:0 -1 0 -1 5 -1 0 -1 0, hwdownload" OUTPUT
22984 • Apply blur:
22986 -i INPUT -vf "hwupload, convolution_opencl=1 1 1 1 1 1 1 1 1:1 1 1 1 1 1 1 1 1:1 1 1 1 1 1 1 1 1:1 1 1 1 1 1 1 1 1:1/9:1/9:1/9:1/9, hwdownload" OUTPUT
22988 • Apply edge enhance:
22990 -i INPUT -vf "hwupload, convolution_opencl=0 0 0 -1 1 0 0 0 0:0 0 0 -1 1 0 0 0 0:0 0 0 -1 1 0 0 0 0:0 0 0 -1 1 0 0 0 0:5:1:1:1:0:128:128:128, hwdownload" OUTPUT
22992 • Apply edge detect:
22994 -i INPUT -vf "hwupload, convolution_opencl=0 1 0 1 -4 1 0 1 0:0 1 0 1 -4 1 0 1 0:0 1 0 1 -4 1 0 1 0:0 1 0 1 -4 1 0 1 0:5:5:5:1:0:128:128:128, hwdownload" OUTPUT
22996 • Apply laplacian edge detector which includes diagonals:
22998 -i INPUT -vf "hwupload, convolution_opencl=1 1 1 1 -8 1 1 1 1:1 1 1 1 -8 1 1 1 1:1 1 1 1 -8 1 1 1 1:1 1 1 1 -8 1 1 1 1:5:5:5:1:0:128:128:0, hwdownload" OUTPUT
23000 • Apply emboss:
23002 -i INPUT -vf "hwupload, convolution_opencl=-2 -1 0 -1 1 1 0 1 2:-2 -1 0 -1 1 1 0 1 2:-2 -1 0 -1 1 1 0 1 2:-2 -1 0 -1 1 1 0 1 2, hwdownload" OUTPUT
23004 erosion_opencl
23006 Apply erosion effect to the video.
23007 This filter replaces the pixel by the local(3x3) minimum.
23009 It accepts the following options:
23011 threshold0
23013 threshold1
23014 threshold2
23015 threshold3
23016 Limit the maximum change for each plane. Range is "[0, 65535]" and
23017 default value is 65535. If 0, plane will remain unchanged.
23018 coordinates
23020 Flag which specifies the pixel to refer to. Range is "[0, 255]"
23021 and default value is 255, i.e. all eight pixels are used.
23022 Flags to local 3x3 coordinates region centered on "x":
23024 1 2 3
23026 4 x 5
23028 6 7 8
23030 Example
23032 • Apply erosion filter with threshold0 set to 30, threshold1 set 40,
23034 threshold2 set to 50 and coordinates set to 231, setting each pixel
23035 of the output to the local minimum between pixels: 1, 2, 3, 6, 7, 8
23036 of the 3x3 region centered on it in the input. If the difference
23037 between input pixel and local minimum is more then threshold of the
23038 corresponding plane, output pixel will be set to input pixel -
23039 threshold of corresponding plane.
23040 -i INPUT -vf "hwupload, erosion_opencl=30:40:50:coordinates=231, hwdownload" OUTPUT
23042 deshake_opencl
23044 Feature-point based video stabilization filter.
23045 The filter accepts the following options:
23047 tripod
23049 Simulates a tripod by preventing any camera movement whatsoever
23050 from the original frame. Defaults to 0.
23051 debug
23053 Whether or not additional debug info should be displayed, both in
23054 the processed output and in the console.
23055 Note that in order to see console debug output you will also need
23057 to pass "-v verbose" to ffmpeg.
23058 Viewing point matches in the output video is only supported for RGB
23060 input.
23061 Defaults to 0.
23063 adaptive_crop
23065 Whether or not to do a tiny bit of cropping at the borders to cut
23066 down on the amount of mirrored pixels.
23067 Defaults to 1.
23069 refine_features
23071 Whether or not feature points should be refined at a sub-pixel
23072 level.
23073 This can be turned off for a slight performance gain at the cost of
23075 precision.
23076 Defaults to 1.
23078 smooth_strength
23080 The strength of the smoothing applied to the camera path from 0.0
23081 to 1.0.
23082 1.0 is the maximum smoothing strength while values less than that
23084 result in less smoothing.
23085 0.0 causes the filter to adaptively choose a smoothing strength on
23087 a per-frame basis.
23088 Defaults to 0.0.
23090 smooth_window_multiplier
23092 Controls the size of the smoothing window (the number of frames
23093 buffered to determine motion information from).
23094 The size of the smoothing window is determined by multiplying the
23096 framerate of the video by this number.
23097 Acceptable values range from 0.1 to 10.0.
23099 Larger values increase the amount of motion data available for
23101 determining how to smooth the camera path, potentially improving
23102 smoothness, but also increase latency and memory usage.
23103 Defaults to 2.0.
23105 Examples
23107 • Stabilize a video with a fixed, medium smoothing strength:
23109 -i INPUT -vf "hwupload, deshake_opencl=smooth_strength=0.5, hwdownload" OUTPUT
23111 • Stabilize a video with debugging (both in console and in rendered
23113 video):
23114 -i INPUT -filter_complex "[0:v]format=rgba, hwupload, deshake_opencl=debug=1, hwdownload, format=rgba, format=yuv420p" -v verbose OUTPUT
23116 dilation_opencl
23118 Apply dilation effect to the video.
23119 This filter replaces the pixel by the local(3x3) maximum.
23121 It accepts the following options:
23123 threshold0
23125 threshold1
23126 threshold2
23127 threshold3
23128 Limit the maximum change for each plane. Range is "[0, 65535]" and
23129 default value is 65535. If 0, plane will remain unchanged.
23130 coordinates
23132 Flag which specifies the pixel to refer to. Range is "[0, 255]"
23133 and default value is 255, i.e. all eight pixels are used.
23134 Flags to local 3x3 coordinates region centered on "x":
23136 1 2 3
23138 4 x 5
23140 6 7 8
23142 Example
23144 • Apply dilation filter with threshold0 set to 30, threshold1 set 40,
23146 threshold2 set to 50 and coordinates set to 231, setting each pixel
23147 of the output to the local maximum between pixels: 1, 2, 3, 6, 7, 8
23148 of the 3x3 region centered on it in the input. If the difference
23149 between input pixel and local maximum is more then threshold of the
23150 corresponding plane, output pixel will be set to input pixel +
23151 threshold of corresponding plane.
23152 -i INPUT -vf "hwupload, dilation_opencl=30:40:50:coordinates=231, hwdownload" OUTPUT
23154 nlmeans_opencl
23156 Non-local Means denoise filter through OpenCL, this filter accepts same
23157 options as nlmeans.
23158 overlay_opencl
23160 Overlay one video on top of another.
23161 It takes two inputs and has one output. The first input is the "main"
23163 video on which the second input is overlaid. This filter requires same
23164 memory layout for all the inputs. So, format conversion may be needed.
23165 The filter accepts the following options:
23167 x Set the x coordinate of the overlaid video on the main video.
23169 Default value is 0.
23170 y Set the y coordinate of the overlaid video on the main video.
23172 Default value is 0.
23173 Examples
23175 • Overlay an image LOGO at the top-left corner of the INPUT video.
23177 Both inputs are yuv420p format.
23178 -i INPUT -i LOGO -filter_complex "[0:v]hwupload[a], [1:v]format=yuv420p, hwupload[b], [a][b]overlay_opencl, hwdownload" OUTPUT
23180 • The inputs have same memory layout for color channels , the overlay
23182 has additional alpha plane, like INPUT is yuv420p, and the LOGO is
23183 yuva420p.
23184 -i INPUT -i LOGO -filter_complex "[0:v]hwupload[a], [1:v]format=yuva420p, hwupload[b], [a][b]overlay_opencl, hwdownload" OUTPUT
23186 pad_opencl
23188 Add paddings to the input image, and place the original input at the
23189 provided x, y coordinates.
23190 It accepts the following options:
23192 width, w
23194 height, h
23195 Specify an expression for the size of the output image with the
23196 paddings added. If the value for width or height is 0, the
23197 corresponding input size is used for the output.
23198 The width expression can reference the value set by the height
23200 expression, and vice versa.
23201 The default value of width and height is 0.
23203 x
23205 y Specify the offsets to place the input image at within the padded
23206 area, with respect to the top/left border of the output image.
23207 The x expression can reference the value set by the y expression,
23209 and vice versa.
23210 The default value of x and y is 0.
23212 If x or y evaluate to a negative number, they'll be changed so the
23214 input image is centered on the padded area.
23215 color
23217 Specify the color of the padded area. For the syntax of this
23218 option, check the "Color" section in the ffmpeg-utils manual.
23219 aspect
23221 Pad to an aspect instead to a resolution.
23222 The value for the width, height, x, and y options are expressions
23224 containing the following constants:
23225 in_w
23227 in_h
23228 The input video width and height.
23229 iw
23231 ih These are the same as in_w and in_h.
23232 out_w
23234 out_h
23235 The output width and height (the size of the padded area), as
23236 specified by the width and height expressions.
23237 ow
23239 oh These are the same as out_w and out_h.
23240 x
23242 y The x and y offsets as specified by the x and y expressions, or NAN
23243 if not yet specified.
23244 a same as iw / ih
23246 sar input sample aspect ratio
23248 dar input display aspect ratio, it is the same as (iw / ih) * sar
23250 prewitt_opencl
23252 Apply the Prewitt operator
23253 (<https://en.wikipedia.org/wiki/Prewitt_operator>) to input video
23254 stream.
23255 The filter accepts the following option:
23257 planes
23259 Set which planes to filter. Default value is 0xf, by which all
23260 planes are processed.
23261 scale
23263 Set value which will be multiplied with filtered result. Range is
23264 "[0.0, 65535]" and default value is 1.0.
23265 delta
23267 Set value which will be added to filtered result. Range is
23268 "[-65535, 65535]" and default value is 0.0.
23269 Example
23271 • Apply the Prewitt operator with scale set to 2 and delta set to 10.
23273 -i INPUT -vf "hwupload, prewitt_opencl=scale=2:delta=10, hwdownload" OUTPUT
23275 program_opencl
23277 Filter video using an OpenCL program.
23278 source
23280 OpenCL program source file.
23281 kernel
23283 Kernel name in program.
23284 inputs
23286 Number of inputs to the filter. Defaults to 1.
23287 size, s
23289 Size of output frames. Defaults to the same as the first input.
23290 The "program_opencl" filter also supports the framesync options.
23292 The program source file must contain a kernel function with the given
23294 name, which will be run once for each plane of the output. Each run on
23295 a plane gets enqueued as a separate 2D global NDRange with one work-
23296 item for each pixel to be generated. The global ID offset for each
23297 work-item is therefore the coordinates of a pixel in the destination
23298 image.
23299 The kernel function needs to take the following arguments:
23301 • Destination image, __write_only image2d_t.
23303 This image will become the output; the kernel should write all of
23305 it.
23306 • Frame index, unsigned int.
23308 This is a counter starting from zero and increasing by one for each
23310 frame.
23311 • Source images, __read_only image2d_t.
23313 These are the most recent images on each input. The kernel may
23315 read from them to generate the output, but they can't be written
23316 to.
23317 Example programs:
23319 • Copy the input to the output (output must be the same size as the
23321 input).
23322 __kernel void copy(__write_only image2d_t destination,
23324 unsigned int index,
23325 __read_only image2d_t source)
23326 {
23327 const sampler_t sampler = CLK_NORMALIZED_COORDS_FALSE;
23328 int2 location = (int2)(get_global_id(0), get_global_id(1));
23330 float4 value = read_imagef(source, sampler, location);
23332 write_imagef(destination, location, value);
23334 }
23335 • Apply a simple transformation, rotating the input by an amount
23337 increasing with the index counter. Pixel values are linearly
23338 interpolated by the sampler, and the output need not have the same
23339 dimensions as the input.
23340 __kernel void rotate_image(__write_only image2d_t dst,
23342 unsigned int index,
23343 __read_only image2d_t src)
23344 {
23345 const sampler_t sampler = (CLK_NORMALIZED_COORDS_FALSE |
23346 CLK_FILTER_LINEAR);
23347 float angle = (float)index / 100.0f;
23349 float2 dst_dim = convert_float2(get_image_dim(dst));
23351 float2 src_dim = convert_float2(get_image_dim(src));
23352 float2 dst_cen = dst_dim / 2.0f;
23354 float2 src_cen = src_dim / 2.0f;
23355 int2 dst_loc = (int2)(get_global_id(0), get_global_id(1));
23357 float2 dst_pos = convert_float2(dst_loc) - dst_cen;
23359 float2 src_pos = {
23360 cos(angle) * dst_pos.x - sin(angle) * dst_pos.y,
23361 sin(angle) * dst_pos.x + cos(angle) * dst_pos.y
23362 };
23363 src_pos = src_pos * src_dim / dst_dim;
23364 float2 src_loc = src_pos + src_cen;
23366 if (src_loc.x < 0.0f || src_loc.y < 0.0f ||
23368 src_loc.x > src_dim.x || src_loc.y > src_dim.y)
23369 write_imagef(dst, dst_loc, 0.5f);
23370 else
23371 write_imagef(dst, dst_loc, read_imagef(src, sampler, src_loc));
23372 }
23373 • Blend two inputs together, with the amount of each input used
23375 varying with the index counter.
23376 __kernel void blend_images(__write_only image2d_t dst,
23378 unsigned int index,
23379 __read_only image2d_t src1,
23380 __read_only image2d_t src2)
23381 {
23382 const sampler_t sampler = (CLK_NORMALIZED_COORDS_FALSE |
23383 CLK_FILTER_LINEAR);
23384 float blend = (cos((float)index / 50.0f) + 1.0f) / 2.0f;
23386 int2 dst_loc = (int2)(get_global_id(0), get_global_id(1));
23388 int2 src1_loc = dst_loc * get_image_dim(src1) / get_image_dim(dst);
23389 int2 src2_loc = dst_loc * get_image_dim(src2) / get_image_dim(dst);
23390 float4 val1 = read_imagef(src1, sampler, src1_loc);
23392 float4 val2 = read_imagef(src2, sampler, src2_loc);
23393 write_imagef(dst, dst_loc, val1 * blend + val2 * (1.0f - blend));
23395 }
23396 remap_opencl
23398 Remap pixels using 2nd: Xmap and 3rd: Ymap input video stream.
23399 Destination pixel at position (X, Y) will be picked from source (x, y)
23401 position where x = Xmap(X, Y) and y = Ymap(X, Y). If mapping values are
23402 out of range, zero value for pixel will be used for destination pixel.
23403 Xmap and Ymap input video streams must be of same dimensions. Output
23405 video stream will have Xmap/Ymap video stream dimensions. Xmap and
23406 Ymap input video streams are 32bit float pixel format, single channel.
23407 interp
23409 Specify interpolation used for remapping of pixels. Allowed values
23410 are "near" and "linear". Default value is "linear".
23411 fill
23413 Specify the color of the unmapped pixels. For the syntax of this
23414 option, check the "Color" section in the ffmpeg-utils manual.
23415 Default color is "black".
23416 roberts_opencl
23418 Apply the Roberts cross operator
23419 (<https://en.wikipedia.org/wiki/Roberts_cross>) to input video stream.
23420 The filter accepts the following option:
23422 planes
23424 Set which planes to filter. Default value is 0xf, by which all
23425 planes are processed.
23426 scale
23428 Set value which will be multiplied with filtered result. Range is
23429 "[0.0, 65535]" and default value is 1.0.
23430 delta
23432 Set value which will be added to filtered result. Range is
23433 "[-65535, 65535]" and default value is 0.0.
23434 Example
23436 • Apply the Roberts cross operator with scale set to 2 and delta set
23438 to 10
23439 -i INPUT -vf "hwupload, roberts_opencl=scale=2:delta=10, hwdownload" OUTPUT
23441 sobel_opencl
23443 Apply the Sobel operator
23444 (<https://en.wikipedia.org/wiki/Sobel_operator>) to input video stream.
23445 The filter accepts the following option:
23447 planes
23449 Set which planes to filter. Default value is 0xf, by which all
23450 planes are processed.
23451 scale
23453 Set value which will be multiplied with filtered result. Range is
23454 "[0.0, 65535]" and default value is 1.0.
23455 delta
23457 Set value which will be added to filtered result. Range is
23458 "[-65535, 65535]" and default value is 0.0.
23459 Example
23461 • Apply sobel operator with scale set to 2 and delta set to 10
23463 -i INPUT -vf "hwupload, sobel_opencl=scale=2:delta=10, hwdownload" OUTPUT
23465 tonemap_opencl
23467 Perform HDR(PQ/HLG) to SDR conversion with tone-mapping.
23468 It accepts the following parameters:
23470 tonemap
23472 Specify the tone-mapping operator to be used. Same as tonemap
23473 option in tonemap.
23474 param
23476 Tune the tone mapping algorithm. same as param option in tonemap.
23477 desat
23479 Apply desaturation for highlights that exceed this level of
23480 brightness. The higher the parameter, the more color information
23481 will be preserved. This setting helps prevent unnaturally blown-out
23482 colors for super-highlights, by (smoothly) turning into white
23483 instead. This makes images feel more natural, at the cost of
23484 reducing information about out-of-range colors.
23485 The default value is 0.5, and the algorithm here is a little
23487 different from the cpu version tonemap currently. A setting of 0.0
23488 disables this option.
23489 threshold
23491 The tonemapping algorithm parameters is fine-tuned per each scene.
23492 And a threshold is used to detect whether the scene has changed or
23493 not. If the distance between the current frame average brightness
23494 and the current running average exceeds a threshold value, we would
23495 re-calculate scene average and peak brightness. The default value
23496 is 0.2.
23497 format
23499 Specify the output pixel format.
23500 Currently supported formats are:
23502 p010
23504 nv12
23505 range, r
23506 Set the output color range.
23507 Possible values are:
23509 tv/mpeg
23511 pc/jpeg
23512 Default is same as input.
23514 primaries, p
23516 Set the output color primaries.
23517 Possible values are:
23519 bt709
23521 bt2020
23522 Default is same as input.
23524 transfer, t
23526 Set the output transfer characteristics.
23527 Possible values are:
23529 bt709
23531 bt2020
23532 Default is bt709.
23534 matrix, m
23536 Set the output colorspace matrix.
23537 Possible value are:
23539 bt709
23541 bt2020
23542 Default is same as input.
23544 Example
23546 • Convert HDR(PQ/HLG) video to bt2020-transfer-characteristic p010
23548 format using linear operator.
23549 -i INPUT -vf "format=p010,hwupload,tonemap_opencl=t=bt2020:tonemap=linear:format=p010,hwdownload,format=p010" OUTPUT
23551 unsharp_opencl
23553 Sharpen or blur the input video.
23554 It accepts the following parameters:
23556 luma_msize_x, lx
23558 Set the luma matrix horizontal size. Range is "[1, 23]" and
23559 default value is 5.
23560 luma_msize_y, ly
23562 Set the luma matrix vertical size. Range is "[1, 23]" and default
23563 value is 5.
23564 luma_amount, la
23566 Set the luma effect strength. Range is "[-10, 10]" and default
23567 value is 1.0.
23568 Negative values will blur the input video, while positive values
23570 will sharpen it, a value of zero will disable the effect.
23571 chroma_msize_x, cx
23573 Set the chroma matrix horizontal size. Range is "[1, 23]" and
23574 default value is 5.
23575 chroma_msize_y, cy
23577 Set the chroma matrix vertical size. Range is "[1, 23]" and
23578 default value is 5.
23579 chroma_amount, ca
23581 Set the chroma effect strength. Range is "[-10, 10]" and default
23582 value is 0.0.
23583 Negative values will blur the input video, while positive values
23585 will sharpen it, a value of zero will disable the effect.
23586 All parameters are optional and default to the equivalent of the string
23588 '5:5:1.0:5:5:0.0'.
23589 Examples
23591 • Apply strong luma sharpen effect:
23593 -i INPUT -vf "hwupload, unsharp_opencl=luma_msize_x=7:luma_msize_y=7:luma_amount=2.5, hwdownload" OUTPUT
23595 • Apply a strong blur of both luma and chroma parameters:
23597 -i INPUT -vf "hwupload, unsharp_opencl=7:7:-2:7:7:-2, hwdownload" OUTPUT
23599 xfade_opencl
23601 Cross fade two videos with custom transition effect by using OpenCL.
23602 It accepts the following options:
23604 transition
23606 Set one of possible transition effects.
23607 custom
23609 Select custom transition effect, the actual transition
23610 description will be picked from source and kernel options.
23611 fade
23613 wipeleft
23614 wiperight
23615 wipeup
23616 wipedown
23617 slideleft
23618 slideright
23619 slideup
23620 slidedown
23621 Default transition is fade.
23622 source
23624 OpenCL program source file for custom transition.
23625 kernel
23627 Set name of kernel to use for custom transition from program source
23628 file.
23629 duration
23631 Set duration of video transition.
23632 offset
23634 Set time of start of transition relative to first video.
23635 The program source file must contain a kernel function with the given
23637 name, which will be run once for each plane of the output. Each run on
23638 a plane gets enqueued as a separate 2D global NDRange with one work-
23639 item for each pixel to be generated. The global ID offset for each
23640 work-item is therefore the coordinates of a pixel in the destination
23641 image.
23642 The kernel function needs to take the following arguments:
23644 • Destination image, __write_only image2d_t.
23646 This image will become the output; the kernel should write all of
23648 it.
23649 • First Source image, __read_only image2d_t. Second Source image,
23651 __read_only image2d_t.
23652 These are the most recent images on each input. The kernel may
23654 read from them to generate the output, but they can't be written
23655 to.
23656 • Transition progress, float. This value is always between 0 and 1
23658 inclusive.
23659 Example programs:
23661 • Apply dots curtain transition effect:
23663 __kernel void blend_images(__write_only image2d_t dst,
23665 __read_only image2d_t src1,
23666 __read_only image2d_t src2,
23667 float progress)
23668 {
23669 const sampler_t sampler = (CLK_NORMALIZED_COORDS_FALSE |
23670 CLK_FILTER_LINEAR);
23671 int2 p = (int2)(get_global_id(0), get_global_id(1));
23672 float2 rp = (float2)(get_global_id(0), get_global_id(1));
23673 float2 dim = (float2)(get_image_dim(src1).x, get_image_dim(src1).y);
23674 rp = rp / dim;
23675 float2 dots = (float2)(20.0, 20.0);
23677 float2 center = (float2)(0,0);
23678 float2 unused;
23679 float4 val1 = read_imagef(src1, sampler, p);
23681 float4 val2 = read_imagef(src2, sampler, p);
23682 bool next = distance(fract(rp * dots, &unused), (float2)(0.5, 0.5)) < (progress / distance(rp, center));
23683 write_imagef(dst, p, next ? val1 : val2);
23685 }
23686
23688 VAAPI Video filters are usually used with VAAPI decoder and VAAPI
23689 encoder. Below is a description of VAAPI video filters.
23690 To enable compilation of these filters you need to configure FFmpeg
23692 with "--enable-vaapi".
23693 To use vaapi filters, you need to setup the vaapi device correctly. For
23695 more information, please read
23696 <https://trac.ffmpeg.org/wiki/Hardware/VAAPI>
23697 overlay_vaapi
23699 Overlay one video on the top of another.
23700 It takes two inputs and has one output. The first input is the "main"
23702 video on which the second input is overlaid.
23703 The filter accepts the following options:
23705 x
23707 y Set expressions for the x and y coordinates of the overlaid video
23708 on the main video.
23709 Default value is "0" for both expressions.
23711 w
23713 h Set expressions for the width and height the overlaid video on the
23714 main video.
23715 Default values are 'overlay_iw' for 'w' and
23717 'overlay_ih*w/overlay_iw' for 'h'.
23718 The expressions can contain the following parameters:
23720 main_w, W
23722 main_h, H
23723 The main input width and height.
23724 overlay_iw
23726 overlay_ih
23727 The overlay input width and height.
23728 overlay_w, w
23730 overlay_h, h
23731 The overlay output width and height.
23732 overlay_x, x
23734 overlay_y, y
23735 Position of the overlay layer inside of main
23736 alpha
23738 Set transparency of overlaid video. Allowed range is 0.0 to 1.0.
23739 Higher value means lower transparency. Default value is 1.0.
23740 eof_action
23742 See framesync.
23743 shortest
23745 See framesync.
23746 repeatlast
23748 See framesync.
23749 This filter also supports the framesync options.
23751 Examples
23753 • Overlay an image LOGO at the top-left corner of the INPUT video.
23755 Both inputs for this filter are yuv420p format.
23756 -i INPUT -i LOGO -filter_complex "[0:v]hwupload[a], [1:v]format=yuv420p, hwupload[b], [a][b]overlay_vaapi" OUTPUT
23758 • Overlay an image LOGO at the offset (200, 100) from the top-left
23760 corner of the INPUT video. The inputs have same memory layout for
23761 color channels, the overlay has additional alpha plane, like INPUT
23762 is yuv420p, and the LOGO is yuva420p.
23763 -i INPUT -i LOGO -filter_complex "[0:v]hwupload[a], [1:v]format=yuva420p, hwupload[b], [a][b]overlay_vaapi=x=200:y=100:w=400:h=300:alpha=1.0, hwdownload, format=nv12" OUTPUT
23765 tonemap_vaapi
23767 Perform HDR(High Dynamic Range) to SDR(Standard Dynamic Range)
23768 conversion with tone-mapping. It maps the dynamic range of HDR10
23769 content to the SDR content. It currently only accepts HDR10 as input.
23770 It accepts the following parameters:
23772 format
23774 Specify the output pixel format.
23775 Currently supported formats are:
23777 p010
23779 nv12
23780 Default is nv12.
23782 primaries, p
23784 Set the output color primaries.
23785 Default is same as input.
23787 transfer, t
23789 Set the output transfer characteristics.
23790 Default is bt709.
23792 matrix, m
23794 Set the output colorspace matrix.
23795 Default is same as input.
23797 Example
23799 • Convert HDR(HDR10) video to bt2020-transfer-characteristic p010
23801 format
23802 tonemap_vaapi=format=p010:t=bt2020-10
23804 hstack_vaapi
23806 Stack input videos horizontally.
23807 This is the VA-API variant of the hstack filter, each input stream may
23809 have different height, this filter will scale down/up each input stream
23810 while keeping the orignal aspect.
23811 It accepts the following options:
23813 inputs
23815 See hstack.
23816 shortest
23818 See hstack.
23819 height
23821 Set height of output. If set to 0, this filter will set height of
23822 output to height of the first input stream. Default value is 0.
23823 vstack_vaapi
23825 Stack input videos vertically.
23826 This is the VA-API variant of the vstack filter, each input stream may
23828 have different width, this filter will scale down/up each input stream
23829 while keeping the orignal aspect.
23830 It accepts the following options:
23832 inputs
23834 See vstack.
23835 shortest
23837 See vstack.
23838 width
23840 Set width of output. If set to 0, this filter will set width of
23841 output to width of the first input stream. Default value is 0.
23842 xstack_vaapi
23844 Stack video inputs into custom layout.
23845 This is the VA-API variant of the xstack filter, each input stream may
23847 have different size, this filter will scale down/up each input stream
23848 to the given output size, or the size of the first input stream.
23849 It accepts the following options:
23851 inputs
23853 See xstack.
23854 shortest
23856 See xstack.
23857 layout
23859 See xstack. Moreover, this permits the user to supply output size
23860 for each input stream.
23861 xstack_vaapi=inputs=4:layout=0_0_1920x1080|0_h0_1920x1080|w0_0_1920x1080|w0_h0_1920x1080
23863 grid
23865 See xstack.
23866 grid_tile_size
23868 Set output size for each input stream when grid is set. If this
23869 option is not set, this filter will set output size by default to
23870 the size of the first input stream. For the syntax of this option,
23871 check the "Video size" section in the ffmpeg-utils manual.
23872 fill
23874 See xstack.
23875
23877 Below is a description of the currently available QSV video filters.
23878 To enable compilation of these filters you need to configure FFmpeg
23880 with "--enable-libmfx" or "--enable-libvpl".
23881 To use QSV filters, you need to setup the QSV device correctly. For
23883 more information, please read
23884 <https://trac.ffmpeg.org/wiki/Hardware/QuickSync>
23885 hstack_qsv
23887 Stack input videos horizontally.
23888 This is the QSV variant of the hstack filter, each input stream may
23890 have different height, this filter will scale down/up each input stream
23891 while keeping the orignal aspect.
23892 It accepts the following options:
23894 inputs
23896 See hstack.
23897 shortest
23899 See hstack.
23900 height
23902 Set height of output. If set to 0, this filter will set height of
23903 output to height of the first input stream. Default value is 0.
23904 vstack_qsv
23906 Stack input videos vertically.
23907 This is the QSV variant of the vstack filter, each input stream may
23909 have different width, this filter will scale down/up each input stream
23910 while keeping the orignal aspect.
23911 It accepts the following options:
23913 inputs
23915 See vstack.
23916 shortest
23918 See vstack.
23919 width
23921 Set width of output. If set to 0, this filter will set width of
23922 output to width of the first input stream. Default value is 0.
23923 xstack_qsv
23925 Stack video inputs into custom layout.
23926 This is the QSV variant of the xstack filter.
23928 It accepts the following options:
23930 inputs
23932 See xstack.
23933 shortest
23935 See xstack.
23936 layout
23938 See xstack. Moreover, this permits the user to supply output size
23939 for each input stream.
23940 xstack_qsv=inputs=4:layout=0_0_1920x1080|0_h0_1920x1080|w0_0_1920x1080|w0_h0_1920x1080
23942 grid
23944 See xstack.
23945 grid_tile_size
23947 Set output size for each input stream when grid is set. If this
23948 option is not set, this filter will set output size by default to
23949 the size of the first input stream. For the syntax of this option,
23950 check the "Video size" section in the ffmpeg-utils manual.
23951 fill
23953 See xstack.
23954
23956 Below is a description of the currently available video sources.
23957 buffer
23959 Buffer video frames, and make them available to the filter chain.
23960 This source is mainly intended for a programmatic use, in particular
23962 through the interface defined in libavfilter/buffersrc.h.
23963 It accepts the following parameters:
23965 video_size
23967 Specify the size (width and height) of the buffered video frames.
23968 For the syntax of this option, check the "Video size" section in
23969 the ffmpeg-utils manual.
23970 width
23972 The input video width.
23973 height
23975 The input video height.
23976 pix_fmt
23978 A string representing the pixel format of the buffered video
23979 frames. It may be a number corresponding to a pixel format, or a
23980 pixel format name.
23981 time_base
23983 Specify the timebase assumed by the timestamps of the buffered
23984 frames.
23985 frame_rate
23987 Specify the frame rate expected for the video stream.
23988 pixel_aspect, sar
23990 The sample (pixel) aspect ratio of the input video.
23991 hw_frames_ctx
23993 When using a hardware pixel format, this should be a reference to
23994 an AVHWFramesContext describing input frames.
23995 For example:
23997 buffer=width=320:height=240:pix_fmt=yuv410p:time_base=1/24:sar=1
23999 will instruct the source to accept video frames with size 320x240 and
24001 with format "yuv410p", assuming 1/24 as the timestamps timebase and
24002 square pixels (1:1 sample aspect ratio). Since the pixel format with
24003 name "yuv410p" corresponds to the number 6 (check the enum
24004 AVPixelFormat definition in libavutil/pixfmt.h), this example
24005 corresponds to:
24006 buffer=size=320x240:pixfmt=6:time_base=1/24:pixel_aspect=1/1
24008 Alternatively, the options can be specified as a flat string, but this
24010 syntax is deprecated:
24011 width:height:pix_fmt:time_base.num:time_base.den:pixel_aspect.num:pixel_aspect.den
24013 cellauto
24015 Create a pattern generated by an elementary cellular automaton.
24016 The initial state of the cellular automaton can be defined through the
24018 filename and pattern options. If such options are not specified an
24019 initial state is created randomly.
24020 At each new frame a new row in the video is filled with the result of
24022 the cellular automaton next generation. The behavior when the whole
24023 frame is filled is defined by the scroll option.
24024 This source accepts the following options:
24026 filename, f
24028 Read the initial cellular automaton state, i.e. the starting row,
24029 from the specified file. In the file, each non-whitespace
24030 character is considered an alive cell, a newline will terminate the
24031 row, and further characters in the file will be ignored.
24032 pattern, p
24034 Read the initial cellular automaton state, i.e. the starting row,
24035 from the specified string.
24036 Each non-whitespace character in the string is considered an alive
24038 cell, a newline will terminate the row, and further characters in
24039 the string will be ignored.
24040 rate, r
24042 Set the video rate, that is the number of frames generated per
24043 second. Default is 25.
24044 random_fill_ratio, ratio
24046 Set the random fill ratio for the initial cellular automaton row.
24047 It is a floating point number value ranging from 0 to 1, defaults
24048 to 1/PHI.
24049 This option is ignored when a file or a pattern is specified.
24051 random_seed, seed
24053 Set the seed for filling randomly the initial row, must be an
24054 integer included between 0 and UINT32_MAX. If not specified, or if
24055 explicitly set to -1, the filter will try to use a good random seed
24056 on a best effort basis.
24057 rule
24059 Set the cellular automaton rule, it is a number ranging from 0 to
24060 255. Default value is 110.
24061 size, s
24063 Set the size of the output video. For the syntax of this option,
24064 check the "Video size" section in the ffmpeg-utils manual.
24065 If filename or pattern is specified, the size is set by default to
24067 the width of the specified initial state row, and the height is set
24068 to width * PHI.
24069 If size is set, it must contain the width of the specified pattern
24071 string, and the specified pattern will be centered in the larger
24072 row.
24073 If a filename or a pattern string is not specified, the size value
24075 defaults to "320x518" (used for a randomly generated initial
24076 state).
24077 scroll
24079 If set to 1, scroll the output upward when all the rows in the
24080 output have been already filled. If set to 0, the new generated row
24081 will be written over the top row just after the bottom row is
24082 filled. Defaults to 1.
24083 start_full, full
24085 If set to 1, completely fill the output with generated rows before
24086 outputting the first frame. This is the default behavior, for
24087 disabling set the value to 0.
24088 stitch
24090 If set to 1, stitch the left and right row edges together. This is
24091 the default behavior, for disabling set the value to 0.
24092 Examples
24094 • Read the initial state from pattern, and specify an output of size
24096 200x400.
24097 cellauto=f=pattern:s=200x400
24099 • Generate a random initial row with a width of 200 cells, with a
24101 fill ratio of 2/3:
24102 cellauto=ratio=2/3:s=200x200
24104 • Create a pattern generated by rule 18 starting by a single alive
24106 cell centered on an initial row with width 100:
24107 cellauto=p=@s=100x400:full=0:rule=18
24109 • Specify a more elaborated initial pattern:
24111 cellauto=p='@@ @ @@':s=100x400:full=0:rule=18
24113 coreimagesrc
24115 Video source generated on GPU using Apple's CoreImage API on OSX.
24116 This video source is a specialized version of the coreimage video
24118 filter. Use a core image generator at the beginning of the applied
24119 filterchain to generate the content.
24120 The coreimagesrc video source accepts the following options:
24122 list_generators
24124 List all available generators along with all their respective
24125 options as well as possible minimum and maximum values along with
24126 the default values.
24127 list_generators=true
24129 size, s
24131 Specify the size of the sourced video. For the syntax of this
24132 option, check the "Video size" section in the ffmpeg-utils manual.
24133 The default value is "320x240".
24134 rate, r
24136 Specify the frame rate of the sourced video, as the number of
24137 frames generated per second. It has to be a string in the format
24138 frame_rate_num/frame_rate_den, an integer number, a floating point
24139 number or a valid video frame rate abbreviation. The default value
24140 is "25".
24141 sar Set the sample aspect ratio of the sourced video.
24143 duration, d
24145 Set the duration of the sourced video. See the Time duration
24146 section in the ffmpeg-utils(1) manual for the accepted syntax.
24147 If not specified, or the expressed duration is negative, the video
24149 is supposed to be generated forever.
24150 Additionally, all options of the coreimage video filter are accepted.
24152 A complete filterchain can be used for further processing of the
24153 generated input without CPU-HOST transfer. See coreimage documentation
24154 and examples for details.
24155 Examples
24157 • Use CIQRCodeGenerator to create a QR code for the FFmpeg homepage,
24159 given as complete and escaped command-line for Apple's standard
24160 bash shell:
24161 ffmpeg -f lavfi -i coreimagesrc=s=100x100:filter=CIQRCodeGenerator@inputMessage=https\\\\\://FFmpeg.org/@inputCorrectionLevel=H -frames:v 1 QRCode.png
24163 This example is equivalent to the QRCode example of coreimage
24165 without the need for a nullsrc video source.
24166 ddagrab
24168 Captures the Windows Desktop via Desktop Duplication API.
24169 The filter exclusively returns D3D11 Hardware Frames, for on-gpu
24171 encoding or processing. So an explicit hwdownload is needed for any
24172 kind of software processing.
24173 It accepts the following options:
24175 output_idx
24177 DXGI Output Index to capture.
24178 Usually corresponds to the index Windows has given the screen minus
24180 one, so it's starting at 0.
24181 Defaults to output 0.
24183 draw_mouse
24185 Whether to draw the mouse cursor.
24186 Defaults to true.
24188 Only affects hardware cursors. If a game or application renders its
24190 own cursor, it'll always be captured.
24191 framerate
24193 Framerate at which the desktop will be captured.
24194 Defaults to 30 FPS.
24196 video_size
24198 Specify the size of the captured video.
24199 Defaults to the full size of the screen.
24201 Cropped from the bottom/right if smaller than screen size.
24203 offset_x
24205 Horizontal offset of the captured video.
24206 offset_y
24208 Vertical offset of the captured video.
24209 output_fmt
24211 Desired filter output format. Defaults to 8 Bit BGRA.
24212 It accepts the following values:
24214 auto
24216 Passes all supported output formats to DDA and returns what DDA
24217 decides to use.
24218 8bit
24220 bgra
24221 8 Bit formats always work, and DDA will convert to them if
24222 neccesary.
24223 10bit
24225 x2bgr10
24226 Filter initialization will fail if 10 bit format is requested
24227 but unavailable.
24228 Examples
24230 Capture primary screen and encode using nvenc:
24232 ffmpeg -f lavfi -i ddagrab -c:v h264_nvenc -cq 18 output.mp4
24234 You can also skip the lavfi device and directly use the filter. Also
24236 demonstrates downloading the frame and encoding with libx264. Explicit
24237 output format specification is required in this case:
24238 ffmpeg -filter_complex ddagrab=output_idx=1:framerate=60,hwdownload,format=bgra -c:v libx264 -crf 18 output.mp4
24240 If you want to capture only a subsection of the desktop, this can be
24242 achieved by specifying a smaller size and its offsets into the screen:
24243 ddagrab=video_size=800x600:offset_x=100:offset_y=100
24245 gradients
24247 Generate several gradients.
24248 size, s
24250 Set frame size. For the syntax of this option, check the "Video
24251 size" section in the ffmpeg-utils manual. Default value is
24252 "640x480".
24253 rate, r
24255 Set frame rate, expressed as number of frames per second. Default
24256 value is "25".
24257 c0, c1, c2, c3, c4, c5, c6, c7
24259 Set 8 colors. Default values for colors is to pick random one.
24260 x0, y0, y0, y1
24262 Set gradient line source and destination points. If negative or out
24263 of range, random ones are picked.
24264 nb_colors, n
24266 Set number of colors to use at once. Allowed range is from 2 to 8.
24267 Default value is 2.
24268 seed
24270 Set seed for picking gradient line points.
24271 duration, d
24273 Set the duration of the sourced video. See the Time duration
24274 section in the ffmpeg-utils(1) manual for the accepted syntax.
24275 If not specified, or the expressed duration is negative, the video
24277 is supposed to be generated forever.
24278 speed
24280 Set speed of gradients rotation.
24281 type, t
24283 Set type of gradients, can be "linear" or "radial" or "circular" or
24284 "spiral".
24285 mandelbrot
24287 Generate a Mandelbrot set fractal, and progressively zoom towards the
24288 point specified with start_x and start_y.
24289 This source accepts the following options:
24291 end_pts
24293 Set the terminal pts value. Default value is 400.
24294 end_scale
24296 Set the terminal scale value. Must be a floating point value.
24297 Default value is 0.3.
24298 inner
24300 Set the inner coloring mode, that is the algorithm used to draw the
24301 Mandelbrot fractal internal region.
24302 It shall assume one of the following values:
24304 black
24306 Set black mode.
24307 convergence
24309 Show time until convergence.
24310 mincol
24312 Set color based on point closest to the origin of the
24313 iterations.
24314 period
24316 Set period mode.
24317 Default value is mincol.
24319 bailout
24321 Set the bailout value. Default value is 10.0.
24322 maxiter
24324 Set the maximum of iterations performed by the rendering algorithm.
24325 Default value is 7189.
24326 outer
24328 Set outer coloring mode. It shall assume one of following values:
24329 iteration_count
24331 Set iteration count mode.
24332 normalized_iteration_count
24334 set normalized iteration count mode.
24335 Default value is normalized_iteration_count.
24337 rate, r
24339 Set frame rate, expressed as number of frames per second. Default
24340 value is "25".
24341 size, s
24343 Set frame size. For the syntax of this option, check the "Video
24344 size" section in the ffmpeg-utils manual. Default value is
24345 "640x480".
24346 start_scale
24348 Set the initial scale value. Default value is 3.0.
24349 start_x
24351 Set the initial x position. Must be a floating point value between
24352 -100 and 100. Default value is
24353 -0.743643887037158704752191506114774.
24354 start_y
24356 Set the initial y position. Must be a floating point value between
24357 -100 and 100. Default value is
24358 -0.131825904205311970493132056385139.
24359 mptestsrc
24361 Generate various test patterns, as generated by the MPlayer test
24362 filter.
24363 The size of the generated video is fixed, and is 256x256. This source
24365 is useful in particular for testing encoding features.
24366 This source accepts the following options:
24368 rate, r
24370 Specify the frame rate of the sourced video, as the number of
24371 frames generated per second. It has to be a string in the format
24372 frame_rate_num/frame_rate_den, an integer number, a floating point
24373 number or a valid video frame rate abbreviation. The default value
24374 is "25".
24375 duration, d
24377 Set the duration of the sourced video. See the Time duration
24378 section in the ffmpeg-utils(1) manual for the accepted syntax.
24379 If not specified, or the expressed duration is negative, the video
24381 is supposed to be generated forever.
24382 test, t
24384 Set the number or the name of the test to perform. Supported tests
24385 are:
24386 dc_luma
24388 dc_chroma
24389 freq_luma
24390 freq_chroma
24391 amp_luma
24392 amp_chroma
24393 cbp
24394 mv
24395 ring1
24396 ring2
24397 all
24398 max_frames, m
24399 Set the maximum number of frames generated for each test,
24400 default value is 30.
24401 Default value is "all", which will cycle through the list of all
24403 tests.
24404 Some examples:
24406 mptestsrc=t=dc_luma
24408 will generate a "dc_luma" test pattern.
24410 frei0r_src
24412 Provide a frei0r source.
24413 To enable compilation of this filter you need to install the frei0r
24415 header and configure FFmpeg with "--enable-frei0r".
24416 This source accepts the following parameters:
24418 size
24420 The size of the video to generate. For the syntax of this option,
24421 check the "Video size" section in the ffmpeg-utils manual.
24422 framerate
24424 The framerate of the generated video. It may be a string of the
24425 form num/den or a frame rate abbreviation.
24426 filter_name
24428 The name to the frei0r source to load. For more information
24429 regarding frei0r and how to set the parameters, read the frei0r
24430 section in the video filters documentation.
24431 filter_params
24433 A '|'-separated list of parameters to pass to the frei0r source.
24434 For example, to generate a frei0r partik0l source with size 200x200 and
24436 frame rate 10 which is overlaid on the overlay filter main input:
24437 frei0r_src=size=200x200:framerate=10:filter_name=partik0l:filter_params=1234 [overlay]; [in][overlay] overlay
24439 life
24441 Generate a life pattern.
24442 This source is based on a generalization of John Conway's life game.
24444 The sourced input represents a life grid, each pixel represents a cell
24446 which can be in one of two possible states, alive or dead. Every cell
24447 interacts with its eight neighbours, which are the cells that are
24448 horizontally, vertically, or diagonally adjacent.
24449 At each interaction the grid evolves according to the adopted rule,
24451 which specifies the number of neighbor alive cells which will make a
24452 cell stay alive or born. The rule option allows one to specify the rule
24453 to adopt.
24454 This source accepts the following options:
24456 filename, f
24458 Set the file from which to read the initial grid state. In the
24459 file, each non-whitespace character is considered an alive cell,
24460 and newline is used to delimit the end of each row.
24461 If this option is not specified, the initial grid is generated
24463 randomly.
24464 rate, r
24466 Set the video rate, that is the number of frames generated per
24467 second. Default is 25.
24468 random_fill_ratio, ratio
24470 Set the random fill ratio for the initial random grid. It is a
24471 floating point number value ranging from 0 to 1, defaults to 1/PHI.
24472 It is ignored when a file is specified.
24473 random_seed, seed
24475 Set the seed for filling the initial random grid, must be an
24476 integer included between 0 and UINT32_MAX. If not specified, or if
24477 explicitly set to -1, the filter will try to use a good random seed
24478 on a best effort basis.
24479 rule
24481 Set the life rule.
24482 A rule can be specified with a code of the kind "SNS/BNB", where NS
24484 and NB are sequences of numbers in the range 0-8, NS specifies the
24485 number of alive neighbor cells which make a live cell stay alive,
24486 and NB the number of alive neighbor cells which make a dead cell to
24487 become alive (i.e. to "born"). "s" and "b" can be used in place of
24488 "S" and "B", respectively.
24489 Alternatively a rule can be specified by an 18-bits integer. The 9
24491 high order bits are used to encode the next cell state if it is
24492 alive for each number of neighbor alive cells, the low order bits
24493 specify the rule for "borning" new cells. Higher order bits encode
24494 for an higher number of neighbor cells. For example the number
24495 6153 = "(12<<9)+9" specifies a stay alive rule of 12 and a born
24496 rule of 9, which corresponds to "S23/B03".
24497 Default value is "S23/B3", which is the original Conway's game of
24499 life rule, and will keep a cell alive if it has 2 or 3 neighbor
24500 alive cells, and will born a new cell if there are three alive
24501 cells around a dead cell.
24502 size, s
24504 Set the size of the output video. For the syntax of this option,
24505 check the "Video size" section in the ffmpeg-utils manual.
24506 If filename is specified, the size is set by default to the same
24508 size of the input file. If size is set, it must contain the size
24509 specified in the input file, and the initial grid defined in that
24510 file is centered in the larger resulting area.
24511 If a filename is not specified, the size value defaults to
24513 "320x240" (used for a randomly generated initial grid).
24514 stitch
24516 If set to 1, stitch the left and right grid edges together, and the
24517 top and bottom edges also. Defaults to 1.
24518 mold
24520 Set cell mold speed. If set, a dead cell will go from death_color
24521 to mold_color with a step of mold. mold can have a value from 0 to
24522 255.
24523 life_color
24525 Set the color of living (or new born) cells.
24526 death_color
24528 Set the color of dead cells. If mold is set, this is the first
24529 color used to represent a dead cell.
24530 mold_color
24532 Set mold color, for definitely dead and moldy cells.
24533 For the syntax of these 3 color options, check the "Color" section
24535 in the ffmpeg-utils manual.
24536 Examples
24538 • Read a grid from pattern, and center it on a grid of size 300x300
24540 pixels:
24541 life=f=pattern:s=300x300
24543 • Generate a random grid of size 200x200, with a fill ratio of 2/3:
24545 life=ratio=2/3:s=200x200
24547 • Specify a custom rule for evolving a randomly generated grid:
24549 life=rule=S14/B34
24551 • Full example with slow death effect (mold) using ffplay:
24553 ffplay -f lavfi life=s=300x200:mold=10:r=60:ratio=0.1:death_color=#C83232:life_color=#00ff00,scale=1200:800:flags=16
24555 allrgb, allyuv, color, colorchart, colorspectrum, haldclutsrc, nullsrc,
24557 pal75bars, pal100bars, rgbtestsrc, smptebars, smptehdbars, testsrc,
24558 testsrc2, yuvtestsrc
24559 The "allrgb" source returns frames of size 4096x4096 of all rgb colors.
24560 The "allyuv" source returns frames of size 4096x4096 of all yuv colors.
24562 The "color" source provides an uniformly colored input.
24564 The "colorchart" source provides a colors checker chart.
24566 The "colorspectrum" source provides a color spectrum input.
24568 The "haldclutsrc" source provides an identity Hald CLUT. See also
24570 haldclut filter.
24571 The "nullsrc" source returns unprocessed video frames. It is mainly
24573 useful to be employed in analysis / debugging tools, or as the source
24574 for filters which ignore the input data.
24575 The "pal75bars" source generates a color bars pattern, based on EBU PAL
24577 recommendations with 75% color levels.
24578 The "pal100bars" source generates a color bars pattern, based on EBU
24580 PAL recommendations with 100% color levels.
24581 The "rgbtestsrc" source generates an RGB test pattern useful for
24583 detecting RGB vs BGR issues. You should see a red, green and blue
24584 stripe from top to bottom.
24585 The "smptebars" source generates a color bars pattern, based on the
24587 SMPTE Engineering Guideline EG 1-1990.
24588 The "smptehdbars" source generates a color bars pattern, based on the
24590 SMPTE RP 219-2002.
24591 The "testsrc" source generates a test video pattern, showing a color
24593 pattern, a scrolling gradient and a timestamp. This is mainly intended
24594 for testing purposes.
24595 The "testsrc2" source is similar to testsrc, but supports more pixel
24597 formats instead of just "rgb24". This allows using it as an input for
24598 other tests without requiring a format conversion.
24599 The "yuvtestsrc" source generates an YUV test pattern. You should see a
24601 y, cb and cr stripe from top to bottom.
24602 The sources accept the following parameters:
24604 level
24606 Specify the level of the Hald CLUT, only available in the
24607 "haldclutsrc" source. A level of "N" generates a picture of "N*N*N"
24608 by "N*N*N" pixels to be used as identity matrix for 3D lookup
24609 tables. Each component is coded on a "1/(N*N)" scale.
24610 color, c
24612 Specify the color of the source, only available in the "color"
24613 source. For the syntax of this option, check the "Color" section in
24614 the ffmpeg-utils manual.
24615 size, s
24617 Specify the size of the sourced video. For the syntax of this
24618 option, check the "Video size" section in the ffmpeg-utils manual.
24619 The default value is "320x240".
24620 This option is not available with the "allrgb", "allyuv", and
24622 "haldclutsrc" filters.
24623 rate, r
24625 Specify the frame rate of the sourced video, as the number of
24626 frames generated per second. It has to be a string in the format
24627 frame_rate_num/frame_rate_den, an integer number, a floating point
24628 number or a valid video frame rate abbreviation. The default value
24629 is "25".
24630 duration, d
24632 Set the duration of the sourced video. See the Time duration
24633 section in the ffmpeg-utils(1) manual for the accepted syntax.
24634 If not specified, or the expressed duration is negative, the video
24636 is supposed to be generated forever.
24637 Since the frame rate is used as time base, all frames including the
24639 last one will have their full duration. If the specified duration
24640 is not a multiple of the frame duration, it will be rounded up.
24641 sar Set the sample aspect ratio of the sourced video.
24643 alpha
24645 Specify the alpha (opacity) of the background, only available in
24646 the "testsrc2" source. The value must be between 0 (fully
24647 transparent) and 255 (fully opaque, the default).
24648 decimals, n
24650 Set the number of decimals to show in the timestamp, only available
24651 in the "testsrc" source.
24652 The displayed timestamp value will correspond to the original
24654 timestamp value multiplied by the power of 10 of the specified
24655 value. Default value is 0.
24656 type
24658 Set the type of the color spectrum, only available in the
24659 "colorspectrum" source. Can be one of the following:
24660 black
24662 white
24663 all
24664 patch_size
24665 Set patch size of single color patch, only available in the
24666 "colorchart" source. Default is "64x64".
24667 preset
24669 Set colorchecker colors preset, only available in the "colorchart"
24670 source.
24671 Available values are:
24673 reference
24675 skintones
24676 Default value is "reference".
24678 Examples
24680 • Generate a video with a duration of 5.3 seconds, with size 176x144
24682 and a frame rate of 10 frames per second:
24683 testsrc=duration=5.3:size=qcif:rate=10
24685 • The following graph description will generate a red source with an
24687 opacity of 0.2, with size "qcif" and a frame rate of 10 frames per
24688 second:
24689 color=c=red@0.2:s=qcif:r=10
24691 • If the input content is to be ignored, "nullsrc" can be used. The
24693 following command generates noise in the luminance plane by
24694 employing the "geq" filter:
24695 nullsrc=s=256x256, geq=random(1)*255:128:128
24697 Commands
24699 The "color" source supports the following commands:
24701 c, color
24703 Set the color of the created image. Accepts the same syntax of the
24704 corresponding color option.
24705 openclsrc
24707 Generate video using an OpenCL program.
24708 source
24710 OpenCL program source file.
24711 kernel
24713 Kernel name in program.
24714 size, s
24716 Size of frames to generate. This must be set.
24717 format
24719 Pixel format to use for the generated frames. This must be set.
24720 rate, r
24722 Number of frames generated every second. Default value is '25'.
24723 For details of how the program loading works, see the program_opencl
24725 filter.
24726 Example programs:
24728 • Generate a colour ramp by setting pixel values from the position of
24730 the pixel in the output image. (Note that this will work with all
24731 pixel formats, but the generated output will not be the same.)
24732 __kernel void ramp(__write_only image2d_t dst,
24734 unsigned int index)
24735 {
24736 int2 loc = (int2)(get_global_id(0), get_global_id(1));
24737 float4 val;
24739 val.xy = val.zw = convert_float2(loc) / convert_float2(get_image_dim(dst));
24740 write_imagef(dst, loc, val);
24742 }
24743 • Generate a Sierpinski carpet pattern, panning by a single pixel
24745 each frame.
24746 __kernel void sierpinski_carpet(__write_only image2d_t dst,
24748 unsigned int index)
24749 {
24750 int2 loc = (int2)(get_global_id(0), get_global_id(1));
24751 float4 value = 0.0f;
24753 int x = loc.x + index;
24754 int y = loc.y + index;
24755 while (x > 0 || y > 0) {
24756 if (x % 3 == 1 && y % 3 == 1) {
24757 value = 1.0f;
24758 break;
24759 }
24760 x /= 3;
24761 y /= 3;
24762 }
24763 write_imagef(dst, loc, value);
24765 }
24766 sierpinski
24768 Generate a Sierpinski carpet/triangle fractal, and randomly pan around.
24769 This source accepts the following options:
24771 size, s
24773 Set frame size. For the syntax of this option, check the "Video
24774 size" section in the ffmpeg-utils manual. Default value is
24775 "640x480".
24776 rate, r
24778 Set frame rate, expressed as number of frames per second. Default
24779 value is "25".
24780 seed
24782 Set seed which is used for random panning.
24783 jump
24785 Set max jump for single pan destination. Allowed range is from 1 to
24786 10000.
24787 type
24789 Set fractal type, can be default "carpet" or "triangle".
24790
24792 Below is a description of the currently available video sinks.
24793 buffersink
24795 Buffer video frames, and make them available to the end of the filter
24796 graph.
24797 This sink is mainly intended for programmatic use, in particular
24799 through the interface defined in libavfilter/buffersink.h or the
24800 options system.
24801 It accepts a pointer to an AVBufferSinkContext structure, which defines
24803 the incoming buffers' formats, to be passed as the opaque parameter to
24804 "avfilter_init_filter" for initialization.
24805 nullsink
24807 Null video sink: do absolutely nothing with the input video. It is
24808 mainly useful as a template and for use in analysis / debugging tools.
24809
24811 Below is a description of the currently available multimedia filters.
24812 a3dscope
24814 Convert input audio to 3d scope video output.
24815 The filter accepts the following options:
24817 rate, r
24819 Set frame rate, expressed as number of frames per second. Default
24820 value is "25".
24821 size, s
24823 Specify the video size for the output. For the syntax of this
24824 option, check the "Video size" section in the ffmpeg-utils manual.
24825 Default value is "hd720".
24826 fov Set the camera field of view. Default is 90 degrees. Allowed range
24828 is from 40 to 150.
24829 roll
24831 Set the camera roll.
24832 pitch
24834 Set the camera pitch.
24835 yaw Set the camera yaw.
24837 xzoom
24839 Set the camera zoom on X-axis.
24840 yzoom
24842 Set the camera zoom on Y-axis.
24843 zzoom
24845 Set the camera zoom on Z-axis.
24846 xpos
24848 Set the camera position on X-axis.
24849 ypos
24851 Set the camera position on Y-axis.
24852 zpos
24854 Set the camera position on Z-axis.
24855 length
24857 Set the length of displayed audio waves in number of frames.
24858 Commands
24860 Filter supports the some above options as commands.
24862 abitscope
24864 Convert input audio to a video output, displaying the audio bit scope.
24865 The filter accepts the following options:
24867 rate, r
24869 Set frame rate, expressed as number of frames per second. Default
24870 value is "25".
24871 size, s
24873 Specify the video size for the output. For the syntax of this
24874 option, check the "Video size" section in the ffmpeg-utils manual.
24875 Default value is "1024x256".
24876 colors
24878 Specify list of colors separated by space or by '|' which will be
24879 used to draw channels. Unrecognized or missing colors will be
24880 replaced by white color.
24881 mode, m
24883 Set output mode. Can be "bars" or "trace". Default is "bars".
24884 adrawgraph
24886 Draw a graph using input audio metadata.
24887 See drawgraph
24889 agraphmonitor
24891 See graphmonitor.
24892 ahistogram
24894 Convert input audio to a video output, displaying the volume histogram.
24895 The filter accepts the following options:
24897 dmode
24899 Specify how histogram is calculated.
24900 It accepts the following values:
24902 single
24904 Use single histogram for all channels.
24905 separate
24907 Use separate histogram for each channel.
24908 Default is "single".
24910 rate, r
24912 Set frame rate, expressed as number of frames per second. Default
24913 value is "25".
24914 size, s
24916 Specify the video size for the output. For the syntax of this
24917 option, check the "Video size" section in the ffmpeg-utils manual.
24918 Default value is "hd720".
24919 scale
24921 Set display scale.
24922 It accepts the following values:
24924 log logarithmic
24926 sqrt
24928 square root
24929 cbrt
24931 cubic root
24932 lin linear
24934 rlog
24936 reverse logarithmic
24937 Default is "log".
24939 ascale
24941 Set amplitude scale.
24942 It accepts the following values:
24944 log logarithmic
24946 lin linear
24948 Default is "log".
24950 acount
24952 Set how much frames to accumulate in histogram. Default is 1.
24953 Setting this to -1 accumulates all frames.
24954 rheight
24956 Set histogram ratio of window height.
24957 slide
24959 Set sonogram sliding.
24960 It accepts the following values:
24962 replace
24964 replace old rows with new ones.
24965 scroll
24967 scroll from top to bottom.
24968 Default is "replace".
24970 hmode
24972 Set histogram mode.
24973 It accepts the following values:
24975 abs Use absolute values of samples.
24977 sign
24979 Use untouched values of samples.
24980 Default is "abs".
24982 aphasemeter
24984 Measures phase of input audio, which is exported as metadata
24985 "lavfi.aphasemeter.phase", representing mean phase of current audio
24986 frame. A video output can also be produced and is enabled by default.
24987 The audio is passed through as first output.
24988 Audio will be rematrixed to stereo if it has a different channel
24990 layout. Phase value is in range "[-1, 1]" where -1 means left and right
24991 channels are completely out of phase and 1 means channels are in phase.
24992 The filter accepts the following options, all related to its video
24994 output:
24995 rate, r
24997 Set the output frame rate. Default value is 25.
24998 size, s
25000 Set the video size for the output. For the syntax of this option,
25001 check the "Video size" section in the ffmpeg-utils manual. Default
25002 value is "800x400".
25003 rc
25005 gc
25006 bc Specify the red, green, blue contrast. Default values are 2, 7 and
25007 1. Allowed range is "[0, 255]".
25008 mpc Set color which will be used for drawing median phase. If color is
25010 "none" which is default, no median phase value will be drawn.
25011 video
25013 Enable video output. Default is enabled.
25014 phasing detection
25016 The filter also detects out of phase and mono sequences in stereo
25018 streams. It logs the sequence start, end and duration when it lasts
25019 longer or as long as the minimum set.
25020 The filter accepts the following options for this detection:
25022 phasing
25024 Enable mono and out of phase detection. Default is disabled.
25025 tolerance, t
25027 Set phase tolerance for mono detection, in amplitude ratio. Default
25028 is 0. Allowed range is "[0, 1]".
25029 angle, a
25031 Set angle threshold for out of phase detection, in degree. Default
25032 is 170. Allowed range is "[90, 180]".
25033 duration, d
25035 Set mono or out of phase duration until notification, expressed in
25036 seconds. Default is 2.
25037 Examples
25039 • Complete example with ffmpeg to detect 1 second of mono with 0.001
25041 phase tolerance:
25042 ffmpeg -i stereo.wav -af aphasemeter=video=0:phasing=1:duration=1:tolerance=0.001 -f null -
25044 avectorscope
25046 Convert input audio to a video output, representing the audio vector
25047 scope.
25048 The filter is used to measure the difference between channels of stereo
25050 audio stream. A monaural signal, consisting of identical left and right
25051 signal, results in straight vertical line. Any stereo separation is
25052 visible as a deviation from this line, creating a Lissajous figure. If
25053 the straight (or deviation from it) but horizontal line appears this
25054 indicates that the left and right channels are out of phase.
25055 The filter accepts the following options:
25057 mode, m
25059 Set the vectorscope mode.
25060 Available values are:
25062 lissajous
25064 Lissajous rotated by 45 degrees.
25065 lissajous_xy
25067 Same as above but not rotated.
25068 polar
25070 Shape resembling half of circle.
25071 Default value is lissajous.
25073 size, s
25075 Set the video size for the output. For the syntax of this option,
25076 check the "Video size" section in the ffmpeg-utils manual. Default
25077 value is "400x400".
25078 rate, r
25080 Set the output frame rate. Default value is 25.
25081 rc
25083 gc
25084 bc
25085 ac Specify the red, green, blue and alpha contrast. Default values are
25086 40, 160, 80 and 255. Allowed range is "[0, 255]".
25087 rf
25089 gf
25090 bf
25091 af Specify the red, green, blue and alpha fade. Default values are 15,
25092 10, 5 and 5. Allowed range is "[0, 255]".
25093 zoom
25095 Set the zoom factor. Default value is 1. Allowed range is "[0,
25096 10]". Values lower than 1 will auto adjust zoom factor to maximal
25097 possible value.
25098 draw
25100 Set the vectorscope drawing mode.
25101 Available values are:
25103 dot Draw dot for each sample.
25105 line
25107 Draw line between previous and current sample.
25108 aaline
25110 Draw anti-aliased line between previous and current sample.
25111 Default value is dot.
25113 scale
25115 Specify amplitude scale of audio samples.
25116 Available values are:
25118 lin Linear.
25120 sqrt
25122 Square root.
25123 cbrt
25125 Cubic root.
25126 log Logarithmic.
25128 swap
25130 Swap left channel axis with right channel axis.
25131 mirror
25133 Mirror axis.
25134 none
25136 No mirror.
25137 x Mirror only x axis.
25139 y Mirror only y axis.
25141 xy Mirror both axis.
25143 Examples
25145 • Complete example using ffplay:
25147 ffplay -f lavfi 'amovie=input.mp3, asplit [a][out1];
25149 [a] avectorscope=zoom=1.3:rc=2:gc=200:bc=10:rf=1:gf=8:bf=7 [out0]'
25150 Commands
25152 This filter supports the all above options as commands except options
25154 "size" and "rate".
25155 bench, abench
25157 Benchmark part of a filtergraph.
25158 The filter accepts the following options:
25160 action
25162 Start or stop a timer.
25163 Available values are:
25165 start
25167 Get the current time, set it as frame metadata (using the key
25168 "lavfi.bench.start_time"), and forward the frame to the next
25169 filter.
25170 stop
25172 Get the current time and fetch the "lavfi.bench.start_time"
25173 metadata from the input frame metadata to get the time
25174 difference. Time difference, average, maximum and minimum time
25175 (respectively "t", "avg", "max" and "min") are then printed.
25176 The timestamps are expressed in seconds.
25177 Examples
25179 • Benchmark selectivecolor filter:
25181 bench=start,selectivecolor=reds=-.2 .12 -.49,bench=stop
25183 concat
25185 Concatenate audio and video streams, joining them together one after
25186 the other.
25187 The filter works on segments of synchronized video and audio streams.
25189 All segments must have the same number of streams of each type, and
25190 that will also be the number of streams at output.
25191 The filter accepts the following options:
25193 n Set the number of segments. Default is 2.
25195 v Set the number of output video streams, that is also the number of
25197 video streams in each segment. Default is 1.
25198 a Set the number of output audio streams, that is also the number of
25200 audio streams in each segment. Default is 0.
25201 unsafe
25203 Activate unsafe mode: do not fail if segments have a different
25204 format.
25205 The filter has v+a outputs: first v video outputs, then a audio
25207 outputs.
25208 There are nx(v+a) inputs: first the inputs for the first segment, in
25210 the same order as the outputs, then the inputs for the second segment,
25211 etc.
25212 Related streams do not always have exactly the same duration, for
25214 various reasons including codec frame size or sloppy authoring. For
25215 that reason, related synchronized streams (e.g. a video and its audio
25216 track) should be concatenated at once. The concat filter will use the
25217 duration of the longest stream in each segment (except the last one),
25218 and if necessary pad shorter audio streams with silence.
25219 For this filter to work correctly, all segments must start at timestamp
25221 0.
25222 All corresponding streams must have the same parameters in all
25224 segments; the filtering system will automatically select a common pixel
25225 format for video streams, and a common sample format, sample rate and
25226 channel layout for audio streams, but other settings, such as
25227 resolution, must be converted explicitly by the user.
25228 Different frame rates are acceptable but will result in variable frame
25230 rate at output; be sure to configure the output file to handle it.
25231 Examples
25233 • Concatenate an opening, an episode and an ending, all in bilingual
25235 version (video in stream 0, audio in streams 1 and 2):
25236 ffmpeg -i opening.mkv -i episode.mkv -i ending.mkv -filter_complex \
25238 '[0:0] [0:1] [0:2] [1:0] [1:1] [1:2] [2:0] [2:1] [2:2]
25239 concat=n=3:v=1:a=2 [v] [a1] [a2]' \
25240 -map '[v]' -map '[a1]' -map '[a2]' output.mkv
25241 • Concatenate two parts, handling audio and video separately, using
25243 the (a)movie sources, and adjusting the resolution:
25244 movie=part1.mp4, scale=512:288 [v1] ; amovie=part1.mp4 [a1] ;
25246 movie=part2.mp4, scale=512:288 [v2] ; amovie=part2.mp4 [a2] ;
25247 [v1] [v2] concat [outv] ; [a1] [a2] concat=v=0:a=1 [outa]
25248 Note that a desync will happen at the stitch if the audio and video
25250 streams do not have exactly the same duration in the first file.
25251 Commands
25253 This filter supports the following commands:
25255 next
25257 Close the current segment and step to the next one
25258 ebur128
25260 EBU R128 scanner filter. This filter takes an audio stream and analyzes
25261 its loudness level. By default, it logs a message at a frequency of
25262 10Hz with the Momentary loudness (identified by "M"), Short-term
25263 loudness ("S"), Integrated loudness ("I") and Loudness Range ("LRA").
25264 The filter can only analyze streams which have sample format is double-
25266 precision floating point. The input stream will be converted to this
25267 specification, if needed. Users may need to insert aformat and/or
25268 aresample filters after this filter to obtain the original parameters.
25269 The filter also has a video output (see the video option) with a real
25271 time graph to observe the loudness evolution. The graphic contains the
25272 logged message mentioned above, so it is not printed anymore when this
25273 option is set, unless the verbose logging is set. The main graphing
25274 area contains the short-term loudness (3 seconds of analysis), and the
25275 gauge on the right is for the momentary loudness (400 milliseconds),
25276 but can optionally be configured to instead display short-term loudness
25277 (see gauge).
25278 The green area marks a +/- 1LU target range around the target loudness
25280 (-23LUFS by default, unless modified through target).
25281 More information about the Loudness Recommendation EBU R128 on
25283 <http://tech.ebu.ch/loudness>.
25284 The filter accepts the following options:
25286 video
25288 Activate the video output. The audio stream is passed unchanged
25289 whether this option is set or no. The video stream will be the
25290 first output stream if activated. Default is 0.
25291 size
25293 Set the video size. This option is for video only. For the syntax
25294 of this option, check the "Video size" section in the ffmpeg-utils
25295 manual. Default and minimum resolution is "640x480".
25296 meter
25298 Set the EBU scale meter. Default is 9. Common values are 9 and 18,
25299 respectively for EBU scale meter +9 and EBU scale meter +18. Any
25300 other integer value between this range is allowed.
25301 metadata
25303 Set metadata injection. If set to 1, the audio input will be
25304 segmented into 100ms output frames, each of them containing various
25305 loudness information in metadata. All the metadata keys are
25306 prefixed with "lavfi.r128.".
25307 Default is 0.
25309 framelog
25311 Force the frame logging level.
25312 Available values are:
25314 quiet
25316 logging disabled
25317 info
25319 information logging level
25320 verbose
25322 verbose logging level
25323 By default, the logging level is set to info. If the video or the
25325 metadata options are set, it switches to verbose.
25326 peak
25328 Set peak mode(s).
25329 Available modes can be cumulated (the option is a "flag" type).
25331 Possible values are:
25332 none
25334 Disable any peak mode (default).
25335 sample
25337 Enable sample-peak mode.
25338 Simple peak mode looking for the higher sample value. It logs a
25340 message for sample-peak (identified by "SPK").
25341 true
25343 Enable true-peak mode.
25344 If enabled, the peak lookup is done on an over-sampled version
25346 of the input stream for better peak accuracy. It logs a message
25347 for true-peak. (identified by "TPK") and true-peak per frame
25348 (identified by "FTPK"). This mode requires a build with
25349 "libswresample".
25350 dualmono
25352 Treat mono input files as "dual mono". If a mono file is intended
25353 for playback on a stereo system, its EBU R128 measurement will be
25354 perceptually incorrect. If set to "true", this option will
25355 compensate for this effect. Multi-channel input files are not
25356 affected by this option.
25357 panlaw
25359 Set a specific pan law to be used for the measurement of dual mono
25360 files. This parameter is optional, and has a default value of
25361 -3.01dB.
25362 target
25364 Set a specific target level (in LUFS) used as relative zero in the
25365 visualization. This parameter is optional and has a default value
25366 of -23LUFS as specified by EBU R128. However, material published
25367 online may prefer a level of -16LUFS (e.g. for use with podcasts or
25368 video platforms).
25369 gauge
25371 Set the value displayed by the gauge. Valid values are "momentary"
25372 and s "shortterm". By default the momentary value will be used, but
25373 in certain scenarios it may be more useful to observe the short
25374 term value instead (e.g. live mixing).
25375 scale
25377 Sets the display scale for the loudness. Valid parameters are
25378 "absolute" (in LUFS) or "relative" (LU) relative to the target.
25379 This only affects the video output, not the summary or continuous
25380 log output.
25381 Examples
25383 • Real-time graph using ffplay, with a EBU scale meter +18:
25385 ffplay -f lavfi -i "amovie=input.mp3,ebur128=video=1:meter=18 [out0][out1]"
25387 • Run an analysis with ffmpeg:
25389 ffmpeg -nostats -i input.mp3 -filter_complex ebur128 -f null -
25391 interleave, ainterleave
25393 Temporally interleave frames from several inputs.
25394 "interleave" works with video inputs, "ainterleave" with audio.
25396 These filters read frames from several inputs and send the oldest
25398 queued frame to the output.
25399 Input streams must have well defined, monotonically increasing frame
25401 timestamp values.
25402 In order to submit one frame to output, these filters need to enqueue
25404 at least one frame for each input, so they cannot work in case one
25405 input is not yet terminated and will not receive incoming frames.
25406 For example consider the case when one input is a "select" filter which
25408 always drops input frames. The "interleave" filter will keep reading
25409 from that input, but it will never be able to send new frames to output
25410 until the input sends an end-of-stream signal.
25411 Also, depending on inputs synchronization, the filters will drop frames
25413 in case one input receives more frames than the other ones, and the
25414 queue is already filled.
25415 These filters accept the following options:
25417 nb_inputs, n
25419 Set the number of different inputs, it is 2 by default.
25420 duration
25422 How to determine the end-of-stream.
25423 longest
25425 The duration of the longest input. (default)
25426 shortest
25428 The duration of the shortest input.
25429 first
25431 The duration of the first input.
25432 Examples
25434 • Interleave frames belonging to different streams using ffmpeg:
25436 ffmpeg -i bambi.avi -i pr0n.mkv -filter_complex "[0:v][1:v] interleave" out.avi
25438 • Add flickering blur effect:
25440 select='if(gt(random(0), 0.2), 1, 2)':n=2 [tmp], boxblur=2:2, [tmp] interleave
25442 latency, alatency
25444 Measure filtering latency.
25445 Report previous filter filtering latency, delay in number of audio
25447 samples for audio filters or number of video frames for video filters.
25448 On end of input stream, filter will report min and max measured latency
25450 for previous running filter in filtergraph.
25451 metadata, ametadata
25453 Manipulate frame metadata.
25454 This filter accepts the following options:
25456 mode
25458 Set mode of operation of the filter.
25459 Can be one of the following:
25461 select
25463 If both "value" and "key" is set, select frames which have such
25464 metadata. If only "key" is set, select every frame that has
25465 such key in metadata.
25466 add Add new metadata "key" and "value". If key is already available
25468 do nothing.
25469 modify
25471 Modify value of already present key.
25472 delete
25474 If "value" is set, delete only keys that have such value.
25475 Otherwise, delete key. If "key" is not set, delete all metadata
25476 values in the frame.
25477 print
25479 Print key and its value if metadata was found. If "key" is not
25480 set print all metadata values available in frame.
25481 key Set key used with all modes. Must be set for all modes except
25483 "print" and "delete".
25484 value
25486 Set metadata value which will be used. This option is mandatory for
25487 "modify" and "add" mode.
25488 function
25490 Which function to use when comparing metadata value and "value".
25491 Can be one of following:
25493 same_str
25495 Values are interpreted as strings, returns true if metadata
25496 value is same as "value".
25497 starts_with
25499 Values are interpreted as strings, returns true if metadata
25500 value starts with the "value" option string.
25501 less
25503 Values are interpreted as floats, returns true if metadata
25504 value is less than "value".
25505 equal
25507 Values are interpreted as floats, returns true if "value" is
25508 equal with metadata value.
25509 greater
25511 Values are interpreted as floats, returns true if metadata
25512 value is greater than "value".
25513 expr
25515 Values are interpreted as floats, returns true if expression
25516 from option "expr" evaluates to true.
25517 ends_with
25519 Values are interpreted as strings, returns true if metadata
25520 value ends with the "value" option string.
25521 expr
25523 Set expression which is used when "function" is set to "expr". The
25524 expression is evaluated through the eval API and can contain the
25525 following constants:
25526 VALUE1, FRAMEVAL
25528 Float representation of "value" from metadata key.
25529 VALUE2, USERVAL
25531 Float representation of "value" as supplied by user in "value"
25532 option.
25533 file
25535 If specified in "print" mode, output is written to the named file.
25536 Instead of plain filename any writable url can be specified.
25537 Filename ``-'' is a shorthand for standard output. If "file" option
25538 is not set, output is written to the log with AV_LOG_INFO loglevel.
25539 direct
25541 Reduces buffering in print mode when output is written to a URL set
25542 using file.
25543 Examples
25545 • Print all metadata values for frames with key
25547 "lavfi.signalstats.YDIF" with values between 0 and 1.
25548 signalstats,metadata=print:key=lavfi.signalstats.YDIF:value=0:function=expr:expr='between(VALUE1,0,1)'
25550 • Print silencedetect output to file metadata.txt.
25552 silencedetect,ametadata=mode=print:file=metadata.txt
25554 • Direct all metadata to a pipe with file descriptor 4.
25556 metadata=mode=print:file='pipe\:4'
25558 perms, aperms
25560 Set read/write permissions for the output frames.
25561 These filters are mainly aimed at developers to test direct path in the
25563 following filter in the filtergraph.
25564 The filters accept the following options:
25566 mode
25568 Select the permissions mode.
25569 It accepts the following values:
25571 none
25573 Do nothing. This is the default.
25574 ro Set all the output frames read-only.
25576 rw Set all the output frames directly writable.
25578 toggle
25580 Make the frame read-only if writable, and writable if read-
25581 only.
25582 random
25584 Set each output frame read-only or writable randomly.
25585 seed
25587 Set the seed for the random mode, must be an integer included
25588 between 0 and "UINT32_MAX". If not specified, or if explicitly set
25589 to -1, the filter will try to use a good random seed on a best
25590 effort basis.
25591 Note: in case of auto-inserted filter between the permission filter and
25593 the following one, the permission might not be received as expected in
25594 that following filter. Inserting a format or aformat filter before the
25595 perms/aperms filter can avoid this problem.
25596 realtime, arealtime
25598 Slow down filtering to match real time approximately.
25599 These filters will pause the filtering for a variable amount of time to
25601 match the output rate with the input timestamps. They are similar to
25602 the re option to "ffmpeg".
25603 They accept the following options:
25605 limit
25607 Time limit for the pauses. Any pause longer than that will be
25608 considered a timestamp discontinuity and reset the timer. Default
25609 is 2 seconds.
25610 speed
25612 Speed factor for processing. The value must be a float larger than
25613 zero. Values larger than 1.0 will result in faster than realtime
25614 processing, smaller will slow processing down. The limit is
25615 automatically adapted accordingly. Default is 1.0.
25616 A processing speed faster than what is possible without these
25618 filters cannot be achieved.
25619 Commands
25621 Both filters supports the all above options as commands.
25623 segment, asegment
25625 Split single input stream into multiple streams.
25626 This filter does opposite of concat filters.
25628 "segment" works on video frames, "asegment" on audio samples.
25630 This filter accepts the following options:
25632 timestamps
25634 Timestamps of output segments separated by '|'. The first segment
25635 will run from the beginning of the input stream. The last segment
25636 will run until the end of the input stream
25637 frames, samples
25639 Exact frame/sample count to split the segments.
25640 In all cases, prefixing an each segment with '+' will make it relative
25642 to the previous segment.
25643 Examples
25645 • Split input audio stream into three output audio streams, starting
25647 at start of input audio stream and storing that in 1st output audio
25648 stream, then following at 60th second and storing than in 2nd
25649 output audio stream, and last after 150th second of input audio
25650 stream store in 3rd output audio stream:
25651 asegment=timestamps="60|150"
25653 select, aselect
25655 Select frames to pass in output.
25656 This filter accepts the following options:
25658 expr, e
25660 Set expression, which is evaluated for each input frame.
25661 If the expression is evaluated to zero, the frame is discarded.
25663 If the evaluation result is negative or NaN, the frame is sent to
25665 the first output; otherwise it is sent to the output with index
25666 "ceil(val)-1", assuming that the input index starts from 0.
25667 For example a value of 1.2 corresponds to the output with index
25669 "ceil(1.2)-1 = 2-1 = 1", that is the second output.
25670 outputs, n
25672 Set the number of outputs. The output to which to send the selected
25673 frame is based on the result of the evaluation. Default value is 1.
25674 The expression can contain the following constants:
25676 n The (sequential) number of the filtered frame, starting from 0.
25678 selected_n
25680 The (sequential) number of the selected frame, starting from 0.
25681 prev_selected_n
25683 The sequential number of the last selected frame. It's NAN if
25684 undefined.
25685 TB The timebase of the input timestamps.
25687 pts The PTS (Presentation TimeStamp) of the filtered frame, expressed
25689 in TB units. It's NAN if undefined.
25690 t The PTS of the filtered frame, expressed in seconds. It's NAN if
25692 undefined.
25693 prev_pts
25695 The PTS of the previously filtered frame. It's NAN if undefined.
25696 prev_selected_pts
25698 The PTS of the last previously filtered frame. It's NAN if
25699 undefined.
25700 prev_selected_t
25702 The PTS of the last previously selected frame, expressed in
25703 seconds. It's NAN if undefined.
25704 start_pts
25706 The first PTS in the stream which is not NAN. It remains NAN if not
25707 found.
25708 start_t
25710 The first PTS, in seconds, in the stream which is not NAN. It
25711 remains NAN if not found.
25712 pict_type (video only)
25714 The type of the filtered frame. It can assume one of the following
25715 values:
25716 I
25718 P
25719 B
25720 S
25721 SI
25722 SP
25723 BI
25724 interlace_type (video only)
25725 The frame interlace type. It can assume one of the following
25726 values:
25727 PROGRESSIVE
25729 The frame is progressive (not interlaced).
25730 TOPFIRST
25732 The frame is top-field-first.
25733 BOTTOMFIRST
25735 The frame is bottom-field-first.
25736 consumed_sample_n (audio only)
25738 the number of selected samples before the current frame
25739 samples_n (audio only)
25741 the number of samples in the current frame
25742 sample_rate (audio only)
25744 the input sample rate
25745 key This is 1 if the filtered frame is a key-frame, 0 otherwise.
25747 pos the position in the file of the filtered frame, -1 if the
25749 information is not available (e.g. for synthetic video)
25750 scene (video only)
25752 value between 0 and 1 to indicate a new scene; a low value reflects
25753 a low probability for the current frame to introduce a new scene,
25754 while a higher value means the current frame is more likely to be
25755 one (see the example below)
25756 concatdec_select
25758 The concat demuxer can select only part of a concat input file by
25759 setting an inpoint and an outpoint, but the output packets may not
25760 be entirely contained in the selected interval. By using this
25761 variable, it is possible to skip frames generated by the concat
25762 demuxer which are not exactly contained in the selected interval.
25763 This works by comparing the frame pts against the
25765 lavf.concat.start_time and the lavf.concat.duration packet metadata
25766 values which are also present in the decoded frames.
25767 The concatdec_select variable is -1 if the frame pts is at least
25769 start_time and either the duration metadata is missing or the frame
25770 pts is less than start_time + duration, 0 otherwise, and NaN if the
25771 start_time metadata is missing.
25772 That basically means that an input frame is selected if its pts is
25774 within the interval set by the concat demuxer.
25775 The default value of the select expression is "1".
25777 Examples
25779 • Select all frames in input:
25781 select
25783 The example above is the same as:
25785 select=1
25787 • Skip all frames:
25789 select=0
25791 • Select only I-frames:
25793 select='eq(pict_type\,I)'
25795 • Select one frame every 100:
25797 select='not(mod(n\,100))'
25799 • Select only frames contained in the 10-20 time interval:
25801 select=between(t\,10\,20)
25803 • Select only I-frames contained in the 10-20 time interval:
25805 select=between(t\,10\,20)*eq(pict_type\,I)
25807 • Select frames with a minimum distance of 10 seconds:
25809 select='isnan(prev_selected_t)+gte(t-prev_selected_t\,10)'
25811 • Use aselect to select only audio frames with samples number > 100:
25813 aselect='gt(samples_n\,100)'
25815 • Create a mosaic of the first scenes:
25817 ffmpeg -i video.avi -vf select='gt(scene\,0.4)',scale=160:120,tile -frames:v 1 preview.png
25819 Comparing scene against a value between 0.3 and 0.5 is generally a
25821 sane choice.
25822 • Send even and odd frames to separate outputs, and compose them:
25824 select=n=2:e='mod(n, 2)+1' [odd][even]; [odd] pad=h=2*ih [tmp]; [tmp][even] overlay=y=h
25826 • Select useful frames from an ffconcat file which is using inpoints
25828 and outpoints but where the source files are not intra frame only.
25829 ffmpeg -copyts -vsync 0 -segment_time_metadata 1 -i input.ffconcat -vf select=concatdec_select -af aselect=concatdec_select output.avi
25831 sendcmd, asendcmd
25833 Send commands to filters in the filtergraph.
25834 These filters read commands to be sent to other filters in the
25836 filtergraph.
25837 "sendcmd" must be inserted between two video filters, "asendcmd" must
25839 be inserted between two audio filters, but apart from that they act the
25840 same way.
25841 The specification of commands can be provided in the filter arguments
25843 with the commands option, or in a file specified by the filename
25844 option.
25845 These filters accept the following options:
25847 commands, c
25849 Set the commands to be read and sent to the other filters.
25850 filename, f
25852 Set the filename of the commands to be read and sent to the other
25853 filters.
25854 Commands syntax
25856 A commands description consists of a sequence of interval
25858 specifications, comprising a list of commands to be executed when a
25859 particular event related to that interval occurs. The occurring event
25860 is typically the current frame time entering or leaving a given time
25861 interval.
25862 An interval is specified by the following syntax:
25864 <START>[-<END>] <COMMANDS>;
25866 The time interval is specified by the START and END times. END is
25868 optional and defaults to the maximum time.
25869 The current frame time is considered within the specified interval if
25871 it is included in the interval [START, END), that is when the time is
25872 greater or equal to START and is lesser than END.
25873 COMMANDS consists of a sequence of one or more command specifications,
25875 separated by ",", relating to that interval. The syntax of a command
25876 specification is given by:
25877 [<FLAGS>] <TARGET> <COMMAND> <ARG>
25879 FLAGS is optional and specifies the type of events relating to the time
25881 interval which enable sending the specified command, and must be a non-
25882 null sequence of identifier flags separated by "+" or "|" and enclosed
25883 between "[" and "]".
25884 The following flags are recognized:
25886 enter
25888 The command is sent when the current frame timestamp enters the
25889 specified interval. In other words, the command is sent when the
25890 previous frame timestamp was not in the given interval, and the
25891 current is.
25892 leave
25894 The command is sent when the current frame timestamp leaves the
25895 specified interval. In other words, the command is sent when the
25896 previous frame timestamp was in the given interval, and the current
25897 is not.
25898 expr
25900 The command ARG is interpreted as expression and result of
25901 expression is passed as ARG.
25902 The expression is evaluated through the eval API and can contain
25904 the following constants:
25905 POS Original position in the file of the frame, or undefined if
25907 undefined for the current frame.
25908 PTS The presentation timestamp in input.
25910 N The count of the input frame for video or audio, starting from
25912 0.
25913 T The time in seconds of the current frame.
25915 TS The start time in seconds of the current command interval.
25917 TE The end time in seconds of the current command interval.
25919 TI The interpolated time of the current command interval, TI = (T
25921 - TS) / (TE - TS).
25922 W The video frame width.
25924 H The video frame height.
25926 If FLAGS is not specified, a default value of "[enter]" is assumed.
25928 TARGET specifies the target of the command, usually the name of the
25930 filter class or a specific filter instance name.
25931 COMMAND specifies the name of the command for the target filter.
25933 ARG is optional and specifies the optional list of argument for the
25935 given COMMAND.
25936 Between one interval specification and another, whitespaces, or
25938 sequences of characters starting with "#" until the end of line, are
25939 ignored and can be used to annotate comments.
25940 A simplified BNF description of the commands specification syntax
25942 follows:
25943 <COMMAND_FLAG> ::= "enter" | "leave"
25945 <COMMAND_FLAGS> ::= <COMMAND_FLAG> [(+|"|")<COMMAND_FLAG>]
25946 <COMMAND> ::= ["[" <COMMAND_FLAGS> "]"] <TARGET> <COMMAND> [<ARG>]
25947 <COMMANDS> ::= <COMMAND> [,<COMMANDS>]
25948 <INTERVAL> ::= <START>[-<END>] <COMMANDS>
25949 <INTERVALS> ::= <INTERVAL>[;<INTERVALS>]
25950 Examples
25952 • Specify audio tempo change at second 4:
25954 asendcmd=c='4.0 atempo tempo 1.5',atempo
25956 • Target a specific filter instance:
25958 asendcmd=c='4.0 atempo@my tempo 1.5',atempo@my
25960 • Specify a list of drawtext and hue commands in a file.
25962 # show text in the interval 5-10
25964 5.0-10.0 [enter] drawtext reinit 'fontfile=FreeSerif.ttf:text=hello world',
25965 [leave] drawtext reinit 'fontfile=FreeSerif.ttf:text=';
25966 # desaturate the image in the interval 15-20
25968 15.0-20.0 [enter] hue s 0,
25969 [enter] drawtext reinit 'fontfile=FreeSerif.ttf:text=nocolor',
25970 [leave] hue s 1,
25971 [leave] drawtext reinit 'fontfile=FreeSerif.ttf:text=color';
25972 # apply an exponential saturation fade-out effect, starting from time 25
25974 25 [enter] hue s exp(25-t)
25975 A filtergraph allowing to read and process the above command list
25977 stored in a file test.cmd, can be specified with:
25978 sendcmd=f=test.cmd,drawtext=fontfile=FreeSerif.ttf:text='',hue
25980 setpts, asetpts
25982 Change the PTS (presentation timestamp) of the input frames.
25983 "setpts" works on video frames, "asetpts" on audio frames.
25985 This filter accepts the following options:
25987 expr
25989 The expression which is evaluated for each frame to construct its
25990 timestamp.
25991 The expression is evaluated through the eval API and can contain the
25993 following constants:
25994 FRAME_RATE, FR
25996 frame rate, only defined for constant frame-rate video
25997 PTS The presentation timestamp in input
25999 N The count of the input frame for video or the number of consumed
26001 samples, not including the current frame for audio, starting from
26002 0.
26003 NB_CONSUMED_SAMPLES
26005 The number of consumed samples, not including the current frame
26006 (only audio)
26007 NB_SAMPLES, S
26009 The number of samples in the current frame (only audio)
26010 SAMPLE_RATE, SR
26012 The audio sample rate.
26013 STARTPTS
26015 The PTS of the first frame.
26016 STARTT
26018 the time in seconds of the first frame
26019 INTERLACED
26021 State whether the current frame is interlaced.
26022 T the time in seconds of the current frame
26024 POS original position in the file of the frame, or undefined if
26026 undefined for the current frame
26027 PREV_INPTS
26029 The previous input PTS.
26030 PREV_INT
26032 previous input time in seconds
26033 PREV_OUTPTS
26035 The previous output PTS.
26036 PREV_OUTT
26038 previous output time in seconds
26039 RTCTIME
26041 The wallclock (RTC) time in microseconds. This is deprecated, use
26042 time(0) instead.
26043 RTCSTART
26045 The wallclock (RTC) time at the start of the movie in microseconds.
26046 TB The timebase of the input timestamps.
26048 Examples
26050 • Start counting PTS from zero
26052 setpts=PTS-STARTPTS
26054 • Apply fast motion effect:
26056 setpts=0.5*PTS
26058 • Apply slow motion effect:
26060 setpts=2.0*PTS
26062 • Set fixed rate of 25 frames per second:
26064 setpts=N/(25*TB)
26066 • Set fixed rate 25 fps with some jitter:
26068 setpts='1/(25*TB) * (N + 0.05 * sin(N*2*PI/25))'
26070 • Apply an offset of 10 seconds to the input PTS:
26072 setpts=PTS+10/TB
26074 • Generate timestamps from a "live source" and rebase onto the
26076 current timebase:
26077 setpts='(RTCTIME - RTCSTART) / (TB * 1000000)'
26079 • Generate timestamps by counting samples:
26081 asetpts=N/SR/TB
26083 setrange
26085 Force color range for the output video frame.
26086 The "setrange" filter marks the color range property for the output
26088 frames. It does not change the input frame, but only sets the
26089 corresponding property, which affects how the frame is treated by
26090 following filters.
26091 The filter accepts the following options:
26093 range
26095 Available values are:
26096 auto
26098 Keep the same color range property.
26099 unspecified, unknown
26101 Set the color range as unspecified.
26102 limited, tv, mpeg
26104 Set the color range as limited.
26105 full, pc, jpeg
26107 Set the color range as full.
26108 settb, asettb
26110 Set the timebase to use for the output frames timestamps. It is mainly
26111 useful for testing timebase configuration.
26112 It accepts the following parameters:
26114 expr, tb
26116 The expression which is evaluated into the output timebase.
26117 The value for tb is an arithmetic expression representing a rational.
26119 The expression can contain the constants "AVTB" (the default timebase),
26120 "intb" (the input timebase) and "sr" (the sample rate, audio only).
26121 Default value is "intb".
26122 Examples
26124 • Set the timebase to 1/25:
26126 settb=expr=1/25
26128 • Set the timebase to 1/10:
26130 settb=expr=0.1
26132 • Set the timebase to 1001/1000:
26134 settb=1+0.001
26136 • Set the timebase to 2*intb:
26138 settb=2*intb
26140 • Set the default timebase value:
26142 settb=AVTB
26144 showcqt
26146 Convert input audio to a video output representing frequency spectrum
26147 logarithmically using Brown-Puckette constant Q transform algorithm
26148 with direct frequency domain coefficient calculation (but the transform
26149 itself is not really constant Q, instead the Q factor is actually
26150 variable/clamped), with musical tone scale, from E0 to D#10.
26151 The filter accepts the following options:
26153 size, s
26155 Specify the video size for the output. It must be even. For the
26156 syntax of this option, check the "Video size" section in the
26157 ffmpeg-utils manual. Default value is "1920x1080".
26158 fps, rate, r
26160 Set the output frame rate. Default value is 25.
26161 bar_h
26163 Set the bargraph height. It must be even. Default value is -1 which
26164 computes the bargraph height automatically.
26165 axis_h
26167 Set the axis height. It must be even. Default value is -1 which
26168 computes the axis height automatically.
26169 sono_h
26171 Set the sonogram height. It must be even. Default value is -1 which
26172 computes the sonogram height automatically.
26173 fullhd
26175 Set the fullhd resolution. This option is deprecated, use size, s
26176 instead. Default value is 1.
26177 sono_v, volume
26179 Specify the sonogram volume expression. It can contain variables:
26180 bar_v
26182 the bar_v evaluated expression
26183 frequency, freq, f
26185 the frequency where it is evaluated
26186 timeclamp, tc
26188 the value of timeclamp option
26189 and functions:
26191 a_weighting(f)
26193 A-weighting of equal loudness
26194 b_weighting(f)
26196 B-weighting of equal loudness
26197 c_weighting(f)
26199 C-weighting of equal loudness.
26200 Default value is 16.
26202 bar_v, volume2
26204 Specify the bargraph volume expression. It can contain variables:
26205 sono_v
26207 the sono_v evaluated expression
26208 frequency, freq, f
26210 the frequency where it is evaluated
26211 timeclamp, tc
26213 the value of timeclamp option
26214 and functions:
26216 a_weighting(f)
26218 A-weighting of equal loudness
26219 b_weighting(f)
26221 B-weighting of equal loudness
26222 c_weighting(f)
26224 C-weighting of equal loudness.
26225 Default value is "sono_v".
26227 sono_g, gamma
26229 Specify the sonogram gamma. Lower gamma makes the spectrum more
26230 contrast, higher gamma makes the spectrum having more range.
26231 Default value is 3. Acceptable range is "[1, 7]".
26232 bar_g, gamma2
26234 Specify the bargraph gamma. Default value is 1. Acceptable range is
26235 "[1, 7]".
26236 bar_t
26238 Specify the bargraph transparency level. Lower value makes the
26239 bargraph sharper. Default value is 1. Acceptable range is "[0,
26240 1]".
26241 timeclamp, tc
26243 Specify the transform timeclamp. At low frequency, there is trade-
26244 off between accuracy in time domain and frequency domain. If
26245 timeclamp is lower, event in time domain is represented more
26246 accurately (such as fast bass drum), otherwise event in frequency
26247 domain is represented more accurately (such as bass guitar).
26248 Acceptable range is "[0.002, 1]". Default value is 0.17.
26249 attack
26251 Set attack time in seconds. The default is 0 (disabled). Otherwise,
26252 it limits future samples by applying asymmetric windowing in time
26253 domain, useful when low latency is required. Accepted range is "[0,
26254 1]".
26255 basefreq
26257 Specify the transform base frequency. Default value is
26258 20.01523126408007475, which is frequency 50 cents below E0.
26259 Acceptable range is "[10, 100000]".
26260 endfreq
26262 Specify the transform end frequency. Default value is
26263 20495.59681441799654, which is frequency 50 cents above D#10.
26264 Acceptable range is "[10, 100000]".
26265 coeffclamp
26267 This option is deprecated and ignored.
26268 tlength
26270 Specify the transform length in time domain. Use this option to
26271 control accuracy trade-off between time domain and frequency domain
26272 at every frequency sample. It can contain variables:
26273 frequency, freq, f
26275 the frequency where it is evaluated
26276 timeclamp, tc
26278 the value of timeclamp option.
26279 Default value is "384*tc/(384+tc*f)".
26281 count
26283 Specify the transform count for every video frame. Default value is
26284 6. Acceptable range is "[1, 30]".
26285 fcount
26287 Specify the transform count for every single pixel. Default value
26288 is 0, which makes it computed automatically. Acceptable range is
26289 "[0, 10]".
26290 fontfile
26292 Specify font file for use with freetype to draw the axis. If not
26293 specified, use embedded font. Note that drawing with font file or
26294 embedded font is not implemented with custom basefreq and endfreq,
26295 use axisfile option instead.
26296 font
26298 Specify fontconfig pattern. This has lower priority than fontfile.
26299 The ":" in the pattern may be replaced by "|" to avoid unnecessary
26300 escaping.
26301 fontcolor
26303 Specify font color expression. This is arithmetic expression that
26304 should return integer value 0xRRGGBB. It can contain variables:
26305 frequency, freq, f
26307 the frequency where it is evaluated
26308 timeclamp, tc
26310 the value of timeclamp option
26311 and functions:
26313 midi(f)
26315 midi number of frequency f, some midi numbers: E0(16), C1(24),
26316 C2(36), A4(69)
26317 r(x), g(x), b(x)
26319 red, green, and blue value of intensity x.
26320 Default value is "st(0, (midi(f)-59.5)/12); st(1,
26322 if(between(ld(0),0,1), 0.5-0.5*cos(2*PI*ld(0)), 0)); r(1-ld(1)) +
26323 b(ld(1))".
26324 axisfile
26326 Specify image file to draw the axis. This option override fontfile
26327 and fontcolor option.
26328 axis, text
26330 Enable/disable drawing text to the axis. If it is set to 0, drawing
26331 to the axis is disabled, ignoring fontfile and axisfile option.
26332 Default value is 1.
26333 csp Set colorspace. The accepted values are:
26335 unspecified
26337 Unspecified (default)
26338 bt709
26340 BT.709
26341 fcc FCC
26343 bt470bg
26345 BT.470BG or BT.601-6 625
26346 smpte170m
26348 SMPTE-170M or BT.601-6 525
26349 smpte240m
26351 SMPTE-240M
26352 bt2020ncl
26354 BT.2020 with non-constant luminance
26355 cscheme
26357 Set spectrogram color scheme. This is list of floating point values
26358 with format "left_r|left_g|left_b|right_r|right_g|right_b". The
26359 default is "1|0.5|0|0|0.5|1".
26360 Examples
26362 • Playing audio while showing the spectrum:
26364 ffplay -f lavfi 'amovie=a.mp3, asplit [a][out1]; [a] showcqt [out0]'
26366 • Same as above, but with frame rate 30 fps:
26368 ffplay -f lavfi 'amovie=a.mp3, asplit [a][out1]; [a] showcqt=fps=30:count=5 [out0]'
26370 • Playing at 1280x720:
26372 ffplay -f lavfi 'amovie=a.mp3, asplit [a][out1]; [a] showcqt=s=1280x720:count=4 [out0]'
26374 • Disable sonogram display:
26376 sono_h=0
26378 • A1 and its harmonics: A1, A2, (near)E3, A3:
26380 ffplay -f lavfi 'aevalsrc=0.1*sin(2*PI*55*t)+0.1*sin(4*PI*55*t)+0.1*sin(6*PI*55*t)+0.1*sin(8*PI*55*t),
26382 asplit[a][out1]; [a] showcqt [out0]'
26383 • Same as above, but with more accuracy in frequency domain:
26385 ffplay -f lavfi 'aevalsrc=0.1*sin(2*PI*55*t)+0.1*sin(4*PI*55*t)+0.1*sin(6*PI*55*t)+0.1*sin(8*PI*55*t),
26387 asplit[a][out1]; [a] showcqt=timeclamp=0.5 [out0]'
26388 • Custom volume:
26390 bar_v=10:sono_v=bar_v*a_weighting(f)
26392 • Custom gamma, now spectrum is linear to the amplitude.
26394 bar_g=2:sono_g=2
26396 • Custom tlength equation:
26398 tc=0.33:tlength='st(0,0.17); 384*tc / (384 / ld(0) + tc*f /(1-ld(0))) + 384*tc / (tc*f / ld(0) + 384 /(1-ld(0)))'
26400 • Custom fontcolor and fontfile, C-note is colored green, others are
26402 colored blue:
26403 fontcolor='if(mod(floor(midi(f)+0.5),12), 0x0000FF, g(1))':fontfile=myfont.ttf
26405 • Custom font using fontconfig:
26407 font='Courier New,Monospace,mono|bold'
26409 • Custom frequency range with custom axis using image file:
26411 axisfile=myaxis.png:basefreq=40:endfreq=10000
26413 showcwt
26415 Convert input audio to video output representing frequency spectrum
26416 using Continuous Wavelet Transform and Morlet wavelet.
26417 The filter accepts the following options:
26419 size, s
26421 Specify the video size for the output. For the syntax of this
26422 option, check the "Video size" section in the ffmpeg-utils manual.
26423 Default value is "640x512".
26424 rate, r
26426 Set the output frame rate. Default value is 25.
26427 scale
26429 Set the frequency scale used. Allowed values are:
26430 linear
26432 log2
26433 bark
26434 mel
26435 erbs
26436 Default value is "linear".
26438 min Set the minimum frequency that will be used in output. Default is
26440 20 Hz.
26441 max Set the maximum frequency that will be used in output. Default is
26443 20000 Hz. The real frequency upper limit depends on input audio's
26444 sample rate and such will be enforced on this value when it is set
26445 to value greater than Nyquist frequency.
26446 logb
26448 Set the logarithmic basis for brightness strength when mapping
26449 calculated magnitude values to pixel values. Allowed range is from
26450 0 to 1. Default value is 0.0001.
26451 deviation
26453 Set the frequency deviation. Lower values than 1 are more
26454 frequency oriented, while higher values than 1 are more time
26455 oriented. Allowed range is from 0 to 10. Default value is 1.
26456 pps Set the number of pixel output per each second in one row. Allowed
26458 range is from 1 to 1024. Default value is 64.
26459 mode
26461 Set the output visual mode. Allowed values are:
26462 magnitude
26464 Show magnitude.
26465 phase
26467 Show only phase.
26468 magphase
26470 Show combination of magnitude and phase. Magnitude is mapped
26471 to brightness and phase to color.
26472 channel
26474 Show unique color per channel magnitude.
26475 stereo
26477 Show unique color per stereo difference.
26478 Default value is "magnitude".
26480 slide
26482 Set the output slide method. Allowed values are:
26483 replace
26485 scroll
26486 frame
26487 direction
26488 Set the direction method for output slide method. Allowed values
26489 are:
26490 lr Direction from left to right.
26492 rl Direction from right to left.
26494 ud Direction from up to down.
26496 du Direction from down to up.
26498 showfreqs
26500 Convert input audio to video output representing the audio power
26501 spectrum. Audio amplitude is on Y-axis while frequency is on X-axis.
26502 The filter accepts the following options:
26504 size, s
26506 Specify size of video. For the syntax of this option, check the
26507 "Video size" section in the ffmpeg-utils manual. Default is
26508 "1024x512".
26509 rate, r
26511 Set video rate. Default is 25.
26512 mode
26514 Set display mode. This set how each frequency bin will be
26515 represented.
26516 It accepts the following values:
26518 line
26520 bar
26521 dot
26522 Default is "bar".
26524 ascale
26526 Set amplitude scale.
26527 It accepts the following values:
26529 lin Linear scale.
26531 sqrt
26533 Square root scale.
26534 cbrt
26536 Cubic root scale.
26537 log Logarithmic scale.
26539 Default is "log".
26541 fscale
26543 Set frequency scale.
26544 It accepts the following values:
26546 lin Linear scale.
26548 log Logarithmic scale.
26550 rlog
26552 Reverse logarithmic scale.
26553 Default is "lin".
26555 win_size
26557 Set window size. Allowed range is from 16 to 65536.
26558 Default is 2048
26560 win_func
26562 Set windowing function.
26563 It accepts the following values:
26565 rect
26567 bartlett
26568 hanning
26569 hamming
26570 blackman
26571 welch
26572 flattop
26573 bharris
26574 bnuttall
26575 bhann
26576 sine
26577 nuttall
26578 lanczos
26579 gauss
26580 tukey
26581 dolph
26582 cauchy
26583 parzen
26584 poisson
26585 bohman
26586 kaiser
26587 Default is "hanning".
26589 overlap
26591 Set window overlap. In range "[0, 1]". Default is 1, which means
26592 optimal overlap for selected window function will be picked.
26593 averaging
26595 Set time averaging. Setting this to 0 will display current maximal
26596 peaks. Default is 1, which means time averaging is disabled.
26597 colors
26599 Specify list of colors separated by space or by '|' which will be
26600 used to draw channel frequencies. Unrecognized or missing colors
26601 will be replaced by white color.
26602 cmode
26604 Set channel display mode.
26605 It accepts the following values:
26607 combined
26609 separate
26610 Default is "combined".
26612 minamp
26614 Set minimum amplitude used in "log" amplitude scaler.
26615 data
26617 Set data display mode.
26618 It accepts the following values:
26620 magnitude
26622 phase
26623 delay
26624 Default is "magnitude".
26626 channels
26628 Set channels to use when processing audio. By default all are
26629 processed.
26630 showspatial
26632 Convert stereo input audio to a video output, representing the spatial
26633 relationship between two channels.
26634 The filter accepts the following options:
26636 size, s
26638 Specify the video size for the output. For the syntax of this
26639 option, check the "Video size" section in the ffmpeg-utils manual.
26640 Default value is "512x512".
26641 win_size
26643 Set window size. Allowed range is from 1024 to 65536. Default size
26644 is 4096.
26645 win_func
26647 Set window function.
26648 It accepts the following values:
26650 rect
26652 bartlett
26653 hann
26654 hanning
26655 hamming
26656 blackman
26657 welch
26658 flattop
26659 bharris
26660 bnuttall
26661 bhann
26662 sine
26663 nuttall
26664 lanczos
26665 gauss
26666 tukey
26667 dolph
26668 cauchy
26669 parzen
26670 poisson
26671 bohman
26672 kaiser
26673 Default value is "hann".
26675 rate, r
26677 Set output framerate.
26678 showspectrum
26680 Convert input audio to a video output, representing the audio frequency
26681 spectrum.
26682 The filter accepts the following options:
26684 size, s
26686 Specify the video size for the output. For the syntax of this
26687 option, check the "Video size" section in the ffmpeg-utils manual.
26688 Default value is "640x512".
26689 slide
26691 Specify how the spectrum should slide along the window.
26692 It accepts the following values:
26694 replace
26696 the samples start again on the left when they reach the right
26697 scroll
26699 the samples scroll from right to left
26700 fullframe
26702 frames are only produced when the samples reach the right
26703 rscroll
26705 the samples scroll from left to right
26706 lreplace
26708 the samples start again on the right when they reach the left
26709 Default value is "replace".
26711 mode
26713 Specify display mode.
26714 It accepts the following values:
26716 combined
26718 all channels are displayed in the same row
26719 separate
26721 all channels are displayed in separate rows
26722 Default value is combined.
26724 color
26726 Specify display color mode.
26727 It accepts the following values:
26729 channel
26731 each channel is displayed in a separate color
26732 intensity
26734 each channel is displayed using the same color scheme
26735 rainbow
26737 each channel is displayed using the rainbow color scheme
26738 moreland
26740 each channel is displayed using the moreland color scheme
26741 nebulae
26743 each channel is displayed using the nebulae color scheme
26744 fire
26746 each channel is displayed using the fire color scheme
26747 fiery
26749 each channel is displayed using the fiery color scheme
26750 fruit
26752 each channel is displayed using the fruit color scheme
26753 cool
26755 each channel is displayed using the cool color scheme
26756 magma
26758 each channel is displayed using the magma color scheme
26759 green
26761 each channel is displayed using the green color scheme
26762 viridis
26764 each channel is displayed using the viridis color scheme
26765 plasma
26767 each channel is displayed using the plasma color scheme
26768 cividis
26770 each channel is displayed using the cividis color scheme
26771 terrain
26773 each channel is displayed using the terrain color scheme
26774 Default value is channel.
26776 scale
26778 Specify scale used for calculating intensity color values.
26779 It accepts the following values:
26781 lin linear
26783 sqrt
26785 square root, default
26786 cbrt
26788 cubic root
26789 log logarithmic
26791 4thrt
26793 4th root
26794 5thrt
26796 5th root
26797 Default value is sqrt.
26799 fscale
26801 Specify frequency scale.
26802 It accepts the following values:
26804 lin linear
26806 log logarithmic
26808 Default value is lin.
26810 saturation
26812 Set saturation modifier for displayed colors. Negative values
26813 provide alternative color scheme. 0 is no saturation at all.
26814 Saturation must be in [-10.0, 10.0] range. Default value is 1.
26815 win_func
26817 Set window function.
26818 It accepts the following values:
26820 rect
26822 bartlett
26823 hann
26824 hanning
26825 hamming
26826 blackman
26827 welch
26828 flattop
26829 bharris
26830 bnuttall
26831 bhann
26832 sine
26833 nuttall
26834 lanczos
26835 gauss
26836 tukey
26837 dolph
26838 cauchy
26839 parzen
26840 poisson
26841 bohman
26842 kaiser
26843 Default value is "hann".
26845 orientation
26847 Set orientation of time vs frequency axis. Can be "vertical" or
26848 "horizontal". Default is "vertical".
26849 overlap
26851 Set ratio of overlap window. Default value is 0. When value is 1
26852 overlap is set to recommended size for specific window function
26853 currently used.
26854 gain
26856 Set scale gain for calculating intensity color values. Default
26857 value is 1.
26858 data
26860 Set which data to display. Can be "magnitude", default or "phase",
26861 or unwrapped phase: "uphase".
26862 rotation
26864 Set color rotation, must be in [-1.0, 1.0] range. Default value is
26865 0.
26866 start
26868 Set start frequency from which to display spectrogram. Default is
26869 0.
26870 stop
26872 Set stop frequency to which to display spectrogram. Default is 0.
26873 fps Set upper frame rate limit. Default is "auto", unlimited.
26875 legend
26877 Draw time and frequency axes and legends. Default is disabled.
26878 drange
26880 Set dynamic range used to calculate intensity color values. Default
26881 is 120 dBFS. Allowed range is from 10 to 200.
26882 limit
26884 Set upper limit of input audio samples volume in dBFS. Default is 0
26885 dBFS. Allowed range is from -100 to 100.
26886 opacity
26888 Set opacity strength when using pixel format output with alpha
26889 component.
26890 The usage is very similar to the showwaves filter; see the examples in
26892 that section.
26893 Examples
26895 • Large window with logarithmic color scaling:
26897 showspectrum=s=1280x480:scale=log
26899 • Complete example for a colored and sliding spectrum per channel
26901 using ffplay:
26902 ffplay -f lavfi 'amovie=input.mp3, asplit [a][out1];
26904 [a] showspectrum=mode=separate:color=intensity:slide=1:scale=cbrt [out0]'
26905 showspectrumpic
26907 Convert input audio to a single video frame, representing the audio
26908 frequency spectrum.
26909 The filter accepts the following options:
26911 size, s
26913 Specify the video size for the output. For the syntax of this
26914 option, check the "Video size" section in the ffmpeg-utils manual.
26915 Default value is "4096x2048".
26916 mode
26918 Specify display mode.
26919 It accepts the following values:
26921 combined
26923 all channels are displayed in the same row
26924 separate
26926 all channels are displayed in separate rows
26927 Default value is combined.
26929 color
26931 Specify display color mode.
26932 It accepts the following values:
26934 channel
26936 each channel is displayed in a separate color
26937 intensity
26939 each channel is displayed using the same color scheme
26940 rainbow
26942 each channel is displayed using the rainbow color scheme
26943 moreland
26945 each channel is displayed using the moreland color scheme
26946 nebulae
26948 each channel is displayed using the nebulae color scheme
26949 fire
26951 each channel is displayed using the fire color scheme
26952 fiery
26954 each channel is displayed using the fiery color scheme
26955 fruit
26957 each channel is displayed using the fruit color scheme
26958 cool
26960 each channel is displayed using the cool color scheme
26961 magma
26963 each channel is displayed using the magma color scheme
26964 green
26966 each channel is displayed using the green color scheme
26967 viridis
26969 each channel is displayed using the viridis color scheme
26970 plasma
26972 each channel is displayed using the plasma color scheme
26973 cividis
26975 each channel is displayed using the cividis color scheme
26976 terrain
26978 each channel is displayed using the terrain color scheme
26979 Default value is intensity.
26981 scale
26983 Specify scale used for calculating intensity color values.
26984 It accepts the following values:
26986 lin linear
26988 sqrt
26990 square root, default
26991 cbrt
26993 cubic root
26994 log logarithmic
26996 4thrt
26998 4th root
26999 5thrt
27001 5th root
27002 Default value is log.
27004 fscale
27006 Specify frequency scale.
27007 It accepts the following values:
27009 lin linear
27011 log logarithmic
27013 Default value is lin.
27015 saturation
27017 Set saturation modifier for displayed colors. Negative values
27018 provide alternative color scheme. 0 is no saturation at all.
27019 Saturation must be in [-10.0, 10.0] range. Default value is 1.
27020 win_func
27022 Set window function.
27023 It accepts the following values:
27025 rect
27027 bartlett
27028 hann
27029 hanning
27030 hamming
27031 blackman
27032 welch
27033 flattop
27034 bharris
27035 bnuttall
27036 bhann
27037 sine
27038 nuttall
27039 lanczos
27040 gauss
27041 tukey
27042 dolph
27043 cauchy
27044 parzen
27045 poisson
27046 bohman
27047 kaiser
27048 Default value is "hann".
27050 orientation
27052 Set orientation of time vs frequency axis. Can be "vertical" or
27053 "horizontal". Default is "vertical".
27054 gain
27056 Set scale gain for calculating intensity color values. Default
27057 value is 1.
27058 legend
27060 Draw time and frequency axes and legends. Default is enabled.
27061 rotation
27063 Set color rotation, must be in [-1.0, 1.0] range. Default value is
27064 0.
27065 start
27067 Set start frequency from which to display spectrogram. Default is
27068 0.
27069 stop
27071 Set stop frequency to which to display spectrogram. Default is 0.
27072 drange
27074 Set dynamic range used to calculate intensity color values. Default
27075 is 120 dBFS. Allowed range is from 10 to 200.
27076 limit
27078 Set upper limit of input audio samples volume in dBFS. Default is 0
27079 dBFS. Allowed range is from -100 to 100.
27080 opacity
27082 Set opacity strength when using pixel format output with alpha
27083 component.
27084 Examples
27086 • Extract an audio spectrogram of a whole audio track in a 1024x1024
27088 picture using ffmpeg:
27089 ffmpeg -i audio.flac -lavfi showspectrumpic=s=1024x1024 spectrogram.png
27091 showvolume
27093 Convert input audio volume to a video output.
27094 The filter accepts the following options:
27096 rate, r
27098 Set video rate.
27099 b Set border width, allowed range is [0, 5]. Default is 1.
27101 w Set channel width, allowed range is [80, 8192]. Default is 400.
27103 h Set channel height, allowed range is [1, 900]. Default is 20.
27105 f Set fade, allowed range is [0, 1]. Default is 0.95.
27107 c Set volume color expression.
27109 The expression can use the following variables:
27111 VOLUME
27113 Current max volume of channel in dB.
27114 PEAK
27116 Current peak.
27117 CHANNEL
27119 Current channel number, starting from 0.
27120 t If set, displays channel names. Default is enabled.
27122 v If set, displays volume values. Default is enabled.
27124 o Set orientation, can be horizontal: "h" or vertical: "v", default
27126 is "h".
27127 s Set step size, allowed range is [0, 5]. Default is 0, which means
27129 step is disabled.
27130 p Set background opacity, allowed range is [0, 1]. Default is 0.
27132 m Set metering mode, can be peak: "p" or rms: "r", default is "p".
27134 ds Set display scale, can be linear: "lin" or log: "log", default is
27136 "lin".
27137 dm In second. If set to > 0., display a line for the max level in the
27139 previous seconds. default is disabled: 0.
27140 dmc The color of the max line. Use when "dm" option is set to > 0.
27142 default is: "orange"
27143 showwaves
27145 Convert input audio to a video output, representing the samples waves.
27146 The filter accepts the following options:
27148 size, s
27150 Specify the video size for the output. For the syntax of this
27151 option, check the "Video size" section in the ffmpeg-utils manual.
27152 Default value is "600x240".
27153 mode
27155 Set display mode.
27156 Available values are:
27158 point
27160 Draw a point for each sample.
27161 line
27163 Draw a vertical line for each sample.
27164 p2p Draw a point for each sample and a line between them.
27166 cline
27168 Draw a centered vertical line for each sample.
27169 Default value is "point".
27171 n Set the number of samples which are printed on the same column. A
27173 larger value will decrease the frame rate. Must be a positive
27174 integer. This option can be set only if the value for rate is not
27175 explicitly specified.
27176 rate, r
27178 Set the (approximate) output frame rate. This is done by setting
27179 the option n. Default value is "25".
27180 split_channels
27182 Set if channels should be drawn separately or overlap. Default
27183 value is 0.
27184 colors
27186 Set colors separated by '|' which are going to be used for drawing
27187 of each channel.
27188 scale
27190 Set amplitude scale.
27191 Available values are:
27193 lin Linear.
27195 log Logarithmic.
27197 sqrt
27199 Square root.
27200 cbrt
27202 Cubic root.
27203 Default is linear.
27205 draw
27207 Set the draw mode. This is mostly useful to set for high n.
27208 Available values are:
27210 scale
27212 Scale pixel values for each drawn sample.
27213 full
27215 Draw every sample directly.
27216 Default value is "scale".
27218 Examples
27220 • Output the input file audio and the corresponding video
27222 representation at the same time:
27223 amovie=a.mp3,asplit[out0],showwaves[out1]
27225 • Create a synthetic signal and show it with showwaves, forcing a
27227 frame rate of 30 frames per second:
27228 aevalsrc=sin(1*2*PI*t)*sin(880*2*PI*t):cos(2*PI*200*t),asplit[out0],showwaves=r=30[out1]
27230 showwavespic
27232 Convert input audio to a single video frame, representing the samples
27233 waves.
27234 The filter accepts the following options:
27236 size, s
27238 Specify the video size for the output. For the syntax of this
27239 option, check the "Video size" section in the ffmpeg-utils manual.
27240 Default value is "600x240".
27241 split_channels
27243 Set if channels should be drawn separately or overlap. Default
27244 value is 0.
27245 colors
27247 Set colors separated by '|' which are going to be used for drawing
27248 of each channel.
27249 scale
27251 Set amplitude scale.
27252 Available values are:
27254 lin Linear.
27256 log Logarithmic.
27258 sqrt
27260 Square root.
27261 cbrt
27263 Cubic root.
27264 Default is linear.
27266 draw
27268 Set the draw mode.
27269 Available values are:
27271 scale
27273 Scale pixel values for each drawn sample.
27274 full
27276 Draw every sample directly.
27277 Default value is "scale".
27279 filter
27281 Set the filter mode.
27282 Available values are:
27284 average
27286 Use average samples values for each drawn sample.
27287 peak
27289 Use peak samples values for each drawn sample.
27290 Default value is "average".
27292 Examples
27294 • Extract a channel split representation of the wave form of a whole
27296 audio track in a 1024x800 picture using ffmpeg:
27297 ffmpeg -i audio.flac -lavfi showwavespic=split_channels=1:s=1024x800 waveform.png
27299 sidedata, asidedata
27301 Delete frame side data, or select frames based on it.
27302 This filter accepts the following options:
27304 mode
27306 Set mode of operation of the filter.
27307 Can be one of the following:
27309 select
27311 Select every frame with side data of "type".
27312 delete
27314 Delete side data of "type". If "type" is not set, delete all
27315 side data in the frame.
27316 type
27318 Set side data type used with all modes. Must be set for "select"
27319 mode. For the list of frame side data types, refer to the
27320 "AVFrameSideDataType" enum in libavutil/frame.h. For example, to
27321 choose "AV_FRAME_DATA_PANSCAN" side data, you must specify
27322 "PANSCAN".
27323 spectrumsynth
27325 Synthesize audio from 2 input video spectrums, first input stream
27326 represents magnitude across time and second represents phase across
27327 time. The filter will transform from frequency domain as displayed in
27328 videos back to time domain as presented in audio output.
27329 This filter is primarily created for reversing processed showspectrum
27331 filter outputs, but can synthesize sound from other spectrograms too.
27332 But in such case results are going to be poor if the phase data is not
27333 available, because in such cases phase data need to be recreated,
27334 usually it's just recreated from random noise. For best results use
27335 gray only output ("channel" color mode in showspectrum filter) and
27336 "log" scale for magnitude video and "lin" scale for phase video. To
27337 produce phase, for 2nd video, use "data" option. Inputs videos should
27338 generally use "fullframe" slide mode as that saves resources needed for
27339 decoding video.
27340 The filter accepts the following options:
27342 sample_rate
27344 Specify sample rate of output audio, the sample rate of audio from
27345 which spectrum was generated may differ.
27346 channels
27348 Set number of channels represented in input video spectrums.
27349 scale
27351 Set scale which was used when generating magnitude input spectrum.
27352 Can be "lin" or "log". Default is "log".
27353 slide
27355 Set slide which was used when generating inputs spectrums. Can be
27356 "replace", "scroll", "fullframe" or "rscroll". Default is
27357 "fullframe".
27358 win_func
27360 Set window function used for resynthesis.
27361 overlap
27363 Set window overlap. In range "[0, 1]". Default is 1, which means
27364 optimal overlap for selected window function will be picked.
27365 orientation
27367 Set orientation of input videos. Can be "vertical" or "horizontal".
27368 Default is "vertical".
27369 Examples
27371 • First create magnitude and phase videos from audio, assuming audio
27373 is stereo with 44100 sample rate, then resynthesize videos back to
27374 audio with spectrumsynth:
27375 ffmpeg -i input.flac -lavfi showspectrum=mode=separate:scale=log:overlap=0.875:color=channel:slide=fullframe:data=magnitude -an -c:v rawvideo magnitude.nut
27377 ffmpeg -i input.flac -lavfi showspectrum=mode=separate:scale=lin:overlap=0.875:color=channel:slide=fullframe:data=phase -an -c:v rawvideo phase.nut
27378 ffmpeg -i magnitude.nut -i phase.nut -lavfi spectrumsynth=channels=2:sample_rate=44100:win_func=hann:overlap=0.875:slide=fullframe output.flac
27379 split, asplit
27381 Split input into several identical outputs.
27382 "asplit" works with audio input, "split" with video.
27384 The filter accepts a single parameter which specifies the number of
27386 outputs. If unspecified, it defaults to 2.
27387 Examples
27389 • Create two separate outputs from the same input:
27391 [in] split [out0][out1]
27393 • To create 3 or more outputs, you need to specify the number of
27395 outputs, like in:
27396 [in] asplit=3 [out0][out1][out2]
27398 • Create two separate outputs from the same input, one cropped and
27400 one padded:
27401 [in] split [splitout1][splitout2];
27403 [splitout1] crop=100:100:0:0 [cropout];
27404 [splitout2] pad=200:200:100:100 [padout];
27405 • Create 5 copies of the input audio with ffmpeg:
27407 ffmpeg -i INPUT -filter_complex asplit=5 OUTPUT
27409 zmq, azmq
27411 Receive commands sent through a libzmq client, and forward them to
27412 filters in the filtergraph.
27413 "zmq" and "azmq" work as a pass-through filters. "zmq" must be inserted
27415 between two video filters, "azmq" between two audio filters. Both are
27416 capable to send messages to any filter type.
27417 To enable these filters you need to install the libzmq library and
27419 headers and configure FFmpeg with "--enable-libzmq".
27420 For more information about libzmq see: <http://www.zeromq.org/>
27422 The "zmq" and "azmq" filters work as a libzmq server, which receives
27424 messages sent through a network interface defined by the bind_address
27425 (or the abbreviation "b") option. Default value of this option is
27426 tcp://localhost:5555. You may want to alter this value to your needs,
27427 but do not forget to escape any ':' signs (see filtergraph escaping).
27428 The received message must be in the form:
27430 <TARGET> <COMMAND> [<ARG>]
27432 TARGET specifies the target of the command, usually the name of the
27434 filter class or a specific filter instance name. The default filter
27435 instance name uses the pattern Parsed_<filter_name>_<index>, but you
27436 can override this by using the filter_name@id syntax (see Filtergraph
27437 syntax).
27438 COMMAND specifies the name of the command for the target filter.
27440 ARG is optional and specifies the optional argument list for the given
27442 COMMAND.
27443 Upon reception, the message is processed and the corresponding command
27445 is injected into the filtergraph. Depending on the result, the filter
27446 will send a reply to the client, adopting the format:
27447 <ERROR_CODE> <ERROR_REASON>
27449 <MESSAGE>
27450 MESSAGE is optional.
27452 Examples
27454 Look at tools/zmqsend for an example of a zmq client which can be used
27456 to send commands processed by these filters.
27457 Consider the following filtergraph generated by ffplay. In this
27459 example the last overlay filter has an instance name. All other filters
27460 will have default instance names.
27461 ffplay -dumpgraph 1 -f lavfi "
27463 color=s=100x100:c=red [l];
27464 color=s=100x100:c=blue [r];
27465 nullsrc=s=200x100, zmq [bg];
27466 [bg][l] overlay [bg+l];
27467 [bg+l][r] overlay@my=x=100 "
27468 To change the color of the left side of the video, the following
27470 command can be used:
27471 echo Parsed_color_0 c yellow | tools/zmqsend
27473 To change the right side:
27475 echo Parsed_color_1 c pink | tools/zmqsend
27477 To change the position of the right side:
27479 echo overlay@my x 150 | tools/zmqsend
27481
27483 Below is a description of the currently available multimedia sources.
27484 amovie
27486 This is the same as movie source, except it selects an audio stream by
27487 default.
27488 avsynctest
27490 Generate an Audio/Video Sync Test.
27491 Generated stream periodically shows flash video frame and emits beep in
27493 audio. Useful to inspect A/V sync issues.
27494 It accepts the following options:
27496 size, s
27498 Set output video size. Default value is "hd720".
27499 framerate, fr
27501 Set output video frame rate. Default value is 30.
27502 samplerate, sr
27504 Set output audio sample rate. Default value is 44100.
27505 amplitude, a
27507 Set output audio beep amplitude. Default value is 0.7.
27508 period, p
27510 Set output audio beep period in seconds. Default value is 3.
27511 delay, dl
27513 Set output video flash delay in number of frames. Default value is
27514 0.
27515 cycle, c
27517 Enable cycling of video delays, by default is disabled.
27518 duration, d
27520 Set stream output duration. By default duration is unlimited.
27521 fg, bg, ag
27523 Set foreground/background/additional color.
27524 movie
27526 Read audio and/or video stream(s) from a movie container.
27527 It accepts the following parameters:
27529 filename
27531 The name of the resource to read (not necessarily a file; it can
27532 also be a device or a stream accessed through some protocol).
27533 format_name, f
27535 Specifies the format assumed for the movie to read, and can be
27536 either the name of a container or an input device. If not
27537 specified, the format is guessed from movie_name or by probing.
27538 seek_point, sp
27540 Specifies the seek point in seconds. The frames will be output
27541 starting from this seek point. The parameter is evaluated with
27542 "av_strtod", so the numerical value may be suffixed by an IS
27543 postfix. The default value is "0".
27544 streams, s
27546 Specifies the streams to read. Several streams can be specified,
27547 separated by "+". The source will then have as many outputs, in the
27548 same order. The syntax is explained in the "Stream specifiers"
27549 section in the ffmpeg manual. Two special names, "dv" and "da"
27550 specify respectively the default (best suited) video and audio
27551 stream. Default is "dv", or "da" if the filter is called as
27552 "amovie".
27553 stream_index, si
27555 Specifies the index of the video stream to read. If the value is
27556 -1, the most suitable video stream will be automatically selected.
27557 The default value is "-1". Deprecated. If the filter is called
27558 "amovie", it will select audio instead of video.
27559 loop
27561 Specifies how many times to read the stream in sequence. If the
27562 value is 0, the stream will be looped infinitely. Default value is
27563 "1".
27564 Note that when the movie is looped the source timestamps are not
27566 changed, so it will generate non monotonically increasing
27567 timestamps.
27568 discontinuity
27570 Specifies the time difference between frames above which the point
27571 is considered a timestamp discontinuity which is removed by
27572 adjusting the later timestamps.
27573 dec_threads
27575 Specifies the number of threads for decoding
27576 format_opts
27578 Specify format options for the opened file. Format options can be
27579 specified as a list of key=value pairs separated by ':'. The
27580 following example shows how to add protocol_whitelist and
27581 protocol_blacklist options:
27582 ffplay -f lavfi
27584 "movie=filename='1.sdp':format_opts='protocol_whitelist=file,rtp,udp\:protocol_blacklist=http'"
27585 It allows overlaying a second video on top of the main input of a
27587 filtergraph, as shown in this graph:
27588 input -----------> deltapts0 --> overlay --> output
27590 ^
27591 |
27592 movie --> scale--> deltapts1 -------+
27593 Examples
27595 • Skip 3.2 seconds from the start of the AVI file in.avi, and overlay
27597 it on top of the input labelled "in":
27598 movie=in.avi:seek_point=3.2, scale=180:-1, setpts=PTS-STARTPTS [over];
27600 [in] setpts=PTS-STARTPTS [main];
27601 [main][over] overlay=16:16 [out]
27602 • Read from a video4linux2 device, and overlay it on top of the input
27604 labelled "in":
27605 movie=/dev/video0:f=video4linux2, scale=180:-1, setpts=PTS-STARTPTS [over];
27607 [in] setpts=PTS-STARTPTS [main];
27608 [main][over] overlay=16:16 [out]
27609 • Read the first video stream and the audio stream with id 0x81 from
27611 dvd.vob; the video is connected to the pad named "video" and the
27612 audio is connected to the pad named "audio":
27613 movie=dvd.vob:s=v:0+#0x81 [video] [audio]
27615 Commands
27617 Both movie and amovie support the following commands:
27619 seek
27621 Perform seek using "av_seek_frame". The syntax is: seek
27622 stream_index|timestamp|flags
27623 • stream_index: If stream_index is -1, a default stream is
27625 selected, and timestamp is automatically converted from
27626 AV_TIME_BASE units to the stream specific time_base.
27627 • timestamp: Timestamp in AVStream.time_base units or, if no
27629 stream is specified, in AV_TIME_BASE units.
27630 • flags: Flags which select direction and seeking mode.
27632 get_duration
27634 Get movie duration in AV_TIME_BASE units.
27635
27637 ffmpeg(1), ffplay(1), ffprobe(1), libavfilter(3)
27638
27640 The FFmpeg developers.
27641 For details about the authorship, see the Git history of the project
27643 (https://git.ffmpeg.org/ffmpeg), e.g. by typing the command git log in
27644 the FFmpeg source directory, or browsing the online repository at
27645 <https://git.ffmpeg.org/ffmpeg>.
27646 Maintainers for the specific components are listed in the file
27648 MAINTAINERS in the source code tree.
27649 FFMPEG-FILTERS(1)