[FFmpeg-trac] #9485(undetermined:new): Video frame drops when composing video from multiple generative video filters

[FFmpeg]

#9485: Video frame drops when composing video from multiple generative video
filters
-------------------------------------+-------------------------------------
             Reporter:  megpoid      |                     Type:  defect
               Status:  new          |                 Priority:  normal
            Component:               |                  Version:  git-
  undetermined                       |  master
             Keywords:               |               Blocked By:
             Blocking:               |  Reproduced by developer:  0
Analyzed by developer:  0            |
-------------------------------------+-------------------------------------
 Summary of the bug:
 When trying to use multiple audio visualizer filters and compose them
 together to a single video stream, some filters will fail to output at the
 requested frame rate and some images coming from the filter will be
 duplicated.

 How to reproduce:
 {{{
 % ffmpeg -v verbose -i input.mp3 -filter_complex
 "[0:a]avectorscope=s=256x824:r=30[a];
 [0:a]showcqt=s=1664x1080:text=0:r=30[b];
 [0:a]showwaves=s=256x256:mode=line:draw=full:r=30[c]; [a][c]vstack[d];
 [b][d]hstack[out]" -map "[out]" -map 0:a -c:a libopus -vcodec libx264
 -preset ultrafast -pix_fmt yuv444p10le -t 60 -crf 23 -r 30 out0.mkv
 2>log0.txt


 }}}
 When watching the output video, between timestamps 00:00.307 and
 00:00:340, or timestamps 00:03:407 and 00:03.440, you can see that showcqt
 filter's output on the left has duplicate frames. Between timestamps
 00:03.874 and 00:03.907, the showwaves filter's output in the bottom-right
 corner also froze for a frame.
 ffmpeg version N-104443-gdd7a61f62f-g783935de18+1
 built on 2021 October 29th

 Pastebin of the error log file: https://pastebin.com/raw/4VdUPrD3



 Known workaround
 This issue can be worked around by first encoding each individual filter
 into a PNG image sequence, then composing them in a second run,  then
 converting to the desired format in a third run, like so:

 {{{
 ffmpeg -i input.mp3 -filter_complex
 "[0:a]avectorscope=s=256x824:r=30[out]" -map "[out]" -t 60 -r 30
 a_%05d.png

 ffmpeg -i input.mp3 -filter_complex
 "[0:a]showcqt=s=1664x1080:text=0:r=30[out]" -map "[out]" -t 60 -r 30
 b_%05d.png

 ffmpeg -i input.mp3 -filter_complex
 "[0:a]showwaves=s=256x256:mode=line:draw=full:r=30[out]" -map "[out]" -t
 60 -r 30 c_%05d.png

 ffmpeg -r 30 -i a_%05d.png -r 30 -i b_%05d.png -r 30 -i c_%05d.png
 -filter_complex "[0:v][2:v]vstack[a]; [1:v][a]hstack[out]" -map "[out]" -t
 60 -r 30 d_%05d.png

 ffmpeg -r 30 -i d_%05d.png -i input.mp3 -acodec libopus -vcodec libx264
 -preset ultrafast -pix_fmt yuv444p10le -t 60 -crf 23 -r 30 output.mkv
 }}}





 I have permission in writing to redistribute the attached audio file as
 long as credit is given: https://soundcloud.com/polyak
-- 
Ticket URL: <https://trac.ffmpeg.org/ticket/9485>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker