Add multilingual model support from ResembleAI/chatterbox into chatterbox‑tts-api

[kikimora12304855]

Hi @travisvn (or the whole team),

I’ve been following your FastAPI-powered REST API for ChatterBox TTS repository and I’m impressed not only by its clean design but also by how well it runs.
Currently, I rely on xtts2 because of Russian‑language support; until recently that was missing from the base chatterbox package—now a multilingual model has been added.

Could you let me know if there are plans to integrate this new ResembleAI/chatterbox multilingual capability into your API? Specifically:

Will it be available soon for use via OpenWebUI (as my client interface)?
What is an estimated timeline?
Are any additional endpoints or configuration changes required?
Having Russian support directly in the chatterbox‑tts-api would greatly improve voice synthesis quality across languages.

Thanks a lot for all your hard work on this project!

Best regards,
Kiki

Link to source:

• https://github.com/resemble-ai/chatterbox
• https://huggingface.co/ResembleAI/chatterbox

[JobJobovich]

Hi. I support and ask to add multilingual support if you have time and desire to do it.

[fengwang]

Multilingual support (0.1.4) requires three more packages:

1. pkuseg (possibly can be replaced with spacy-pkuseg if no luck to build it normally)
2. safetensors (no problem to install)
3. pykakasi (no problem to install)
   I also see it not difficult to provide a language_id to the multilingual interface: https://github.com/resemble-ai/chatterbox/blob/1b5ae50585c8fa4d30ca6a7b304d787b4f802a42/multilingual_app.py#L180

[kikimora12304855]

Hi @travisvn,

I just wanted to extend my sincere thanks for implementing multi-language support in your API! With the addition of Russian (and other languages) through the ResembleAI/chatterbox integration, I can now fully switch from xtts2 to your solution without any hassle.

This is a huge improvement—now I can seamlessly use OpenWebUI for voice synthesis in Russian, without needing extra workarounds. Your attention to detail and dedication to making the project more accessible for global users truly made a difference.

Thank you for your hard work, thoughtful implementation, and for making this project even better. Keep up the amazing work—you’re building something truly special!

Best regards,
Kiki

[fengwang]

@travisvn I noticed _is_multilingual is initialized in an async function

chatterbox-tts-api/app/core/tts_model.py

Line 83 in 19d3621

use_multilingual = Config.USE_MULTILINGUAL_MODEL

It could happen that, when a user is adding a voice, the model is not yet initialized, and function is_multilingual()

chatterbox-tts-api/app/core/tts_model.py

Line 157 in 19d3621

def is_multilingual():

returns None.