Skip to content

Comment on The Quite OK Audio Format for Fast, Lossy Compressionparent

Comments

Doesnt Opus (speex?) have some low CPU settings?

It does, and I've tried tweaking that, but the performance difference isn't very significant.

I appear to be able to get maybe 30% better performance -- pretty nice, but not nearly big enough especially on low end servers.

I'm not sure much gets better latency than Opus but LyraV2 seemed interesting https://opensource.googleblog.com/2022/09/lyra-v2-a-better-f...

Could be an option, but we take high audio quality as a point of pride and encode in Opus 128k by default. Audio doesn't only include speech but also any sound effects, media present in-world, etc.

But that might be an interesting experiment. Right now the low cpu usage/high quality/faily high bandwidth usage category is something we're looking to have an option for.

Lyra is a speech-only codec, so it's apples and oranges to compare with Opus for general-purpose audio compression.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.