thanks!
not really the place, but is there some AV1 comparison chart between vanilla {libsvtav1,libaom-av1,librav1e} ffmpeg, SVT-AV1-Essential and, let's say, av1an (for same machine, same quality target)?
I don't see the relevance of FFmpeg vs Av1an here. Anything that cannot work with proper scene change detection should simply be disregarded. The encoders are applications first, why test them any other way than standalone? FFmpeg is just extra hassle. Both are extra overhead.
Testing has been conducted over the years by various community members on our discord servers, or in my case publicly with the blog post series on the codec wiki. The data is available, and as much as everything isn't retested every other month, most testing stays relevant with time so it's worth checking.
Rav1e is irrelevant nowadays, aomenc is cumbersome to use and only ever useful in specific considerations. Most encoders don't use SVT-AV1 by coincidence, it is both the most convenient, most flexible and encoder with the most fidelity potential of the bunch. That's without even talking about encoder forks like SVT-AV1-Essential. Anyway, charts typically put aomenc and SVT-AV1 approximately side by side, but metrics have never reflected visual quality the way a human experiences it. Even the state of the art in recent metrics is questionable.
All in all, that question is a bit tricky and not trivial to answer or interpret.
Comments - 3
Albirew
NekoTrix (uploader)
Albirew