Post

Ruairi (rc55)
@rc55.com

Loving seeing how fast machine learning on audio is progressing things - great quality speech under 1kbps is real, and good quality music under 2.5kbps is a reality as well. That's a CD of music on a floppy disk...I can't believe this hasn't had more coverage! github.com/SonyCSLParis/codicodec

github.comGitHub - SonyCSLParis/codicodec: Encode and decode audio samples to/from continuous and discrete compressed representations!Encode and decode audio samples to/from continuous and discrete compressed representations! - SonyCSLParis/codicodec

0 reposts0 likes

Ruairi (rc55)
@rc55.com

Before anyone asks it is GPU heavy, the weights are 600MB also; but we're already seeing device side models going mainstream so this will be the standard approach in the future as classic codec and machine learning continue to work together. Also the quality is outstanding **for the bitrate**.