Is this could be used for audio synthesis?

[MonolithFoundation]

For instance, LLM out produce snac tokens, and decode into audio?

[itsliupeng]

FYI. https://github.com/gpt-omni/mini-omni uses snac codec to generate audio.

[MonolithFoundation]

Thanks for the hint, how about Chinese?

[MrWaterZhou]

https://github.com/MrWaterZhou/viitor-voice

I tried, and it works well :)

[MonolithFoundation]

Woo, does it support Madrian and Japanese?

[MrWaterZhou]

Woo, does it support Madrian and Japanese?

Not yet, but we are working on Madrian and will release it soon.

[MrWaterZhou]

Woo, does it support Madrian and Japanese?

Our Chinese model has been updated—feel free to give it a try!
https://github.com/viitor-ai/viitor-voice/tree/main

FYI, we’ve noticed that the 24kHz SNAC model doesn’t perform well on Chinese audio, especially with higher-pitched samples. We’re currently experimenting with fine-tuning the decoder using vocos, and so far, the results look promising.