Concat with a single input stream is changing the number of frames in the output stream

[abingham]

I have two complex filter graphs which are identical except that one includes a concat=n=1 edge. I expected them to produce output videos with the same number of frames, but what I find is that the one with the 'concat' has one fewer frame.

Here are the commands:

ffmpeg -v 9 -loglevel 99 -f lavfi -r 30 -i "color=c=yellow:s=1920x1080" -f lavfi -r 30 -i "color=c=red:s=1920x1080" -filter_complex "[1:v]trim=end_frame=1[s0];[s0]concat=n=1[s1];[s1]tpad=start=1732[s2];[0:v][s2]overlay=eof_action=repeat:shortest=True[s3]" -map "[s3]" video_1732.mp4 -y 
ffmpeg -v 9 -loglevel 99 -f lavfi -r 30 -i "color=c=yellow:s=1920x1080" -f lavfi -r 30 -i "color=c=red:s=1920x1080" -filter_complex "[1:v]trim=end_frame=1[s0];[s0]tpad=start=1732[s1];[0:v][s1]overlay=eof_action=repeat:shortest=True[s2]" -map "[s2]" video_1733.mp4 -y

On several systems I've checked, the first graph writes a file with 1732 frames while the second produces 1733 frames. I've primarily tested this on an M1 macbook with this ffmpeg:

ffmpeg version N-111575-g74474a551b-tessus  https://evermeet.cx/ffmpeg/  Copyright (c) 2000-2023 the FFmpeg developers
built with Apple clang version 11.0.0 (clang-1100.0.33.17)
configuration: --cc=/usr/bin/clang --prefix=/opt/ffmpeg --extra-version=tessus --enable-avisynth --enable-fontconfig --enable-gpl --enable-libaom --enable-libass --enable-libbluray --enable-libdav1d --enable-libfreetype --enable-libgsm --enable-libmodplug --enable-libmp3lame --enable-libmysofa --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenh264 --enable-libopenjpeg --enable-libopus --enable-librubberband --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libspeex --enable-libtheora --enable-libtwolame --enable-libvidstab --enable-libvmaf --enable-libvo-amrwbenc --enable-libvorbis --enable-libvpx --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxavs --enable-libxvid --enable-libzimg --enable-libzmq --enable-libzvbi --enable-version3 --pkg-config-flags=--static --disable-ffplay
libavutil      58. 14.100 / 58. 14.100
libavcodec     60. 22.100 / 60. 22.100
libavformat    60. 10.100 / 60. 10.100
libavdevice    60.  2.101 / 60.  2.101
libavfilter     9.  8.102 /  9.  8.102
libswscale      7.  3.100 /  7.  3.100
libswresample   4. 11.100 /  4. 11.100
libpostproc    57.  2.100 / 57.  2.100

I've attached the full output for each command.

[abingham]

The output from the first ffmpeg command, the one which produces a video with 1732 frames.

[abingham]

The output from the second ffmpeg command, the one which produces a video with 1733 frames.

[abingham]

This is a greatly distilled example from a video compositing system I'm writing. Internally, it assumes that concat with one (or more) inputs will produce a stream with a number of frames equal to the sum of the input frames. When this assumption was violated (as demonstrated in this issue), it caused some problems.

Perhaps my assumption about concat is incorrect. If so, I'd appreciate some guidance (docs, etc.) on what to expect from concat so that I can compensate for its behavior. My sense is that my assumption is indeed incorrect, and that I should expect rounding issues from concat (e.g. because it converts between frames numbers and timestamps or something).

Or perhaps this is a bug in concat. I'm not really sure.