mr44er
mr44er
Mhm, I see it this way: without in-depth look in 15/18 cases it gives better VMAF and file size aside, that's likely the price to pay for better quality around...
Other sample, Evil Dead remastered UHD, 0:42 - 0:58 (camera flying over swamp), very grainy `-ss 0:42 -to 0:58 -vcodec ffv1 -vf "crop=2878:2160:480:0" -an -sn quelle15sec.mkv` ``` ab-av1 crf-search -i...
Just noticed they use and explain it (but not why it's not default) for example 2 and 3: https://gitlab.com/AOMediaCodec/SVT-AV1/-/blob/master/Docs/Ffmpeg.md
I don't know the differences between all of these DBs with this use case, but I'm up to testing and reporting how it behaves! :)
VMAF doesn't 'see' the synth grain, it only sees the denoised frames and this will result in a lower VMAF score overall. Some detailed info: https://norkin.org/research/film_grain/index.html https://medium.com/@nasirhemed/a-quick-overview-of-video-compression-and-av1-29dffbdb5cc4 The really tricky...
> It is disabled by default upstream (svt-av1) so won't affect encodes unless explicitly opting in. This behaviour seems correct for ab-av1. Yes, it is correct. I have to revise...
> I don't know how you came to this conclusion, or if you meant to say something else. No, that's exactly what I meant. Grain is just metadata (for svt-av1...
Too bad that he didn't go into more details, but ok, calculate VMAF with grain off is a start.
I like to roll that up again to get some opinions and missing pieces. -> Grain off for VMAF calculation as the guy from Netflix suggests, but what next if...
I would vote for a simple default. ffmpeg would eat a (real) .avi when it was just renamed to .mp4 or .mkv, so why would media detection be important? Also...