reconstructing audio from tokens :)

hubertsiuzdak / snac

Multi-Scale Neural Audio Codec (SNAC) compresses audio into discrete codes at a low bitrate

https://hubertsiuzdak.github.io/snac/

MIT License

439 stars 26 forks source link

Closed christophschuhmann closed 7 months ago

christophschuhmann commented 7 months ago

Hi Hubert,

I already posted this on Discord, but you seem on vacation and since i am super excited, I just post it here as well. :)

christophschuhmann commented 7 months ago

TL;DR: If I have "codes" , a list of 4 tensors for the 32k hz model, how d i get audio_hat? :)

hubertsiuzdak commented 7 months ago

now supported with:

audio_hat = model.decode(codes)