training code and scale up to more languages

[ChuxiJ]

Congratulations on your work! The quality of the music you generate is impressive, and there are many methods and ideas worth learning from. Thank you for your contributions to open-source text-to-music research.

I have some questions about Jamify

1. Is there any training code available?
2. How to train a duration predictor to handle lyrics in more languages?
3. In addition to lyrics, if we want to add descriptions of styles, events, or instruments within the lyrics prompt, how should we do that?

[soujanyaporia]

Thank you!

1. Training code will be released soon. @xhhhhang is busy with some urgent matters, but hopefully he will be able to push commits soon.
2. Please stay tuned for Jam-1. The version we released is a very early version and we will release a much better version soon.
3. You are allowed to provide styles. Check out the space: https://huggingface.co/spaces/declare-lab/JAM

Best,
Soujanya

[xhhhhang]

Hi! Thanks for your interest in JAM.

The training code is now released, please check the latest Github page.

Best,
Renhang