Skip to content

Comment on StreamVC: Real-Time Low-Latency Voice Conversionparent

Comments

Once again we see evidence that AI-for-all is not bottlenecked by research but by the physical limitations of compute infrastructure.

I wouldn't say that when the application is a strung-up Python Frankenstein monster (not to be too demeaning to the author).

More efficient architectures are possible. It's bottlenecked by research.

AboutSource Built by g1lg1l

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