Google Releases libaom 3.7 AV1 Encoder With More Optimizations

[Quackdoc]

Is rav1e actively developed any longer?

yes, it's just all the devs are kind of taking a break, Rav1e still has the greatest potential IMO, but as far as I know, only one dev is getting paid to work on it and it's not in a serious capacity. I myself still use it when I need the highest fidelity and consistency, or when I simply don't want to mess with stupid settings and run the risk of the encoder crashing. (libaom is still crash happy) it is by far the slowest encoder however. and doesn't have great size:efficiency outside of the very high fidelity encodes.

on the other hand, if you want really well fine tuned grain, it's the only encoder that doesn't break with grav1synth constantly. so that could be a nice feature I guess.

I wonder how they compare in performance.

Use libaom for most encodes, It gets the best quality:size ratio across the board when using aom-lavish when not, it looses to rav1e at high fidelity, (but this may have changed this update, I havent tested it yet) svtav1 still wins on anything that needs a relatively fast encoder, but it also now has some impressive quality:size ratio, so for any "normie encode" (IE. more then 5fps) svtav1 will probably be better assuming you arent using av1an. for rav1e read above

[OneTimeShot]

Google Bard invented an entire branch of Physics, complete with nonexistent papers allegedly written by real Physicists in a semi-related field for me. It's the ultimate 1984 scenario!

[Yalok]

Please do not punish yourself or others by using AI agents to do "research". They are not capable of it.

Well, I was just superficially curious and wasn't really doing any kind of research.

[linuxgeex]

Google bard says nvenc is the fastest av1 encoder, but turns out there are many more aspects that are important besides speed: https://g.co/bard/share/edccad9a6de7

Please do not punish yourself or others by using AI agents to do "research". They are not capable of it. They are at best capable of saving you some time by collecting the public apocrypha, flavoured with hallucinations, which is better than no starting place at all when you undertake your own research. And then, do it right. First and foremost: never trust google, never trust an AI agent, always question whether you are satisfying your own confirmation bias through cherry-picking - always look for the merit of an opinion which disagrees with your own.

They really need to start teaching this in school, and the world in general needs to look on AI agent search results with extreme scepticism, if not contempt.

[Yalok]

Google bard says nvenc is the fastest av1 encoder, but turns out there are many more aspects that are important besides speed: https://g.co/bard/share/edccad9a6de7

[avis]

I wonder how they compare in performance.

https://forum.doom9.org/forumdisplay.php?f=84

[Yalok]

Three open source encoders (libaom, svt-av1, rav1e) and quite a few of closed source ones.

I wonder how they compare in performance.

[Ranguvar]

Is rav1e actively developed any longer?

appears so, 0.6.6 was tagged May 16

https://github.com/xiph/rav1e/commits/master

[JEBjames]

Michael

Typo

"and speed-ups ad memory" should be "and speed-ups and memory"

[Errinwright]

Three open source encoders (libaom, svt-av1, rav1e) and quite a few of closed source ones.

Is rav1e actively developed any longer?