Comparing videos with different color ranges. about video-compare HOT 8 CLOSED

I have confirmed that the filter works. Thank you.

from video-compare.

Hi @c0494133d4,

I've not been able to recreate the problem. Are you using the same FFmpeg libraries for both ffplay and video-compare?

from video-compare.

@jonfryd Please look into the 0.1s sample (GoPro -> HandBrake x265)

https://github.com/pixop/video-compare/assets/41767537/5cba3d42-e584-40b6-bb7a-f841dd217b4f
https://github.com/pixop/video-compare/assets/41767537/3a66e230-5d54-4514-86e6-847ac6932c33

They looks exactly the same within one player (QuickTime / ffplay / IINA), while the latter one is lightly darker than the first one in video-compare.

Out of topic problem: they looks different across different players.

from video-compare.

Thanks for submitting a short sample of both videos.

I'm not sure what the problem is to be honest. The videos seem to be correctly displayed with the same brightness level on my 2017 iMac running the 20230807-github version of video-compare.

from video-compare.

Same observation on my Lenovo laptop with Windows 10, whether using the 20230807-github version or today's release of video-compare. The brightness level appears identical for both videos as far as I can tell.

It's a bit of a mystery to me why they appear different on your system, @c0494133d4. Have you discovered any clues yourself?

from video-compare.

No clues yet. Here are screenshots:

I am using the latest version of video-compare and ffmpeg from homebrew on a P3-gammut display (sRGB icc has also been tested).

from video-compare.

Thanks for sharing those screenshots. Clearly there is a difference in brightness level between the two videos when decoded on your system (but not on mine). You are using the standard FFmpeg HEVC software decoder (hevc), right?

from video-compare.

Hey @c0494133d4,

I'm able to reproduce the brightness problem by using the latest release and the videotoolbox HW decoder on video A:

./video-compare --left-hwaccel videotoolbox -d videoA.mp4 videoB.mp4

However, in this case it's actually video B (yuv420p) which is brighter than video A (yuvj420p) which contradicts your description of the issue above.

The problem disappears by converting the pixel format to e.g. yuv422p. This might work as a workaround for now:

./video-compare --left-hwaccel videotoolbox -l format=yuv422p -d videoA.mp4 videoB.mp4

With software decoding both videos still look exactly the same on my system.

Assuming the workaround works for you, as well, do you mind if we close this issue for now?

from video-compare.