Just some synthetic videos with some minor color changes.
From here
See the Makefile file for details.
We got some clips (from
test-videos.co.uk and
h264info.com), and ran the
tools/h264nal tool on them to make sure it does not crash.
Then, we compared the results with those produced by ffmpeg BSF mode.
The process (per-video) is as follows:
# convert mp4 to Annex-B format
$ ffmpeg -y -i clip.mp4 -codec:v copy clip.mp4.264
# get bsf data
$ ffmpeg -i clip.mp4.264 -c:v copy -bsf:v trace_headers -f null - >& clip.mp4.264.bsf
# get h264nal parsed data
$ ./h264nal/build/tools/h264nal --noas-one-line --add-offset --add-length --add-parsed-length clip.mp4.264 > clip.mp4.264.txt
And then we do manual comparison of clip.mp4.264.bsf and clip.mp4.264.txt.
Idea is to test the most common FOSS encoders (x264, OpenH264, and others), using different encoder values, and compare the results with those of ffmpeg BSF mode.