ADD A NEW LANGUAGE

[picolo100]

i've tested it and so far is it really good, you did a great job however this need to be always checked as if unchecked it struggle to understand word it seem, anyway i wanted to know if you can train french language? and if it is possible how long would it take to be finished and available?

[Labmem-Zhouyx]

Thanks for your feedback! We're planning to add multilingual support to VoxCPM, including French, pending data preparation. Aiming for a Q1 2026 release. We'll keep you updated :)

[picolo100]

Thanks for your feedback! We're planning to add multilingual support to VoxCPM, including French, pending data preparation. Aiming for a Q1 2026 release. We'll keep you updated :)

thx , by the way is there any plateform where i can connect fastly with staff or collaborators like you, discord, telegram, wechat? im a stress tester, intermediate comfyui user so i need to have faster responsive reaction etc, my question is btw is it right now possible to clone the entire voice like to you use it permanently, as i wanna have the voice ready and push my mic to say anything than the voice i trained will take the exact thing i say, same tone, etc but says it with the voice identity, like 11labs etc not tryin promot just to explain.

[Labmem-Zhouyx]

We've included a WeChat group QR code on Readme for faster communication, and you're also welcome to ask any questions directly in GitHub issues.

Regarding custom voice cloning, it is currently possible to constrain the model to a specific voice by 1) using a fixed prompt speech clip in a zero-shot way, or 2) applying our fine-tuning scripts like LoRA to train yourself voice model, especially if you have several minutes of training data available.

Let us know if you have more concerns :)

[picolo100]

We've included a WeChat group QR code on Readme for faster communication, and you're also welcome to ask any questions directly in GitHub issues.

Regarding custom voice cloning, it is currently possible to constrain the model to a specific voice by 1) using a fixed prompt speech clip in a zero-shot way, or 2) applying our fine-tuning scripts like LoRA to train yourself voice model, especially if you have several minutes of training data available.

Let us know if you have more concerns :)

thank for answering brother, may i know if this is the correct groupe right?, as anyone didn't answer i want to be sure i've not mistaken the group as it's not talkative. by the way can you provide a step by step or even a brief explanation to be clear on how i need to proceed to do that? im fast learning np.

[Labmem-Zhouyx]

Glad that you found the right group! Feel free to ask any questions about VoxCPM in this group—most members are Chinese, but you're very welcome to post in English too!

For fine-tuning, you can follow our step-by-step guide here: https://github.com/OpenBMB/VoxCPM/blob/main/docs/finetune.md It also includes instructions for using lora_webui, thanks to the contribution by @Ayin1412.

Happy building, and let us know if you run into anything.

[park-onezero]

Do you have any plan to support Korean language?

[Ayin1412]

Do you have any plan to support Korean language?

LoRA can do quite well, but still some flaws on punctuations tested cloning the voice of Angela from korean game Lobotomy Corporation by Project Moon

[picolo100]

Do you have any plan to support Korean language?

LoRA can do quite well, but still some flaws on punctuations tested cloning the voice of Angela from korean game Lobotomy Corporation by Project Moon

brother i responded in the other github issue field : can you check it please im waiting since 72 hours https://github.com/OpenBMB/VoxCPM/issues/114#issuecomment-3640767342

[xxoospring]

run lora inference will report error when i modify the lora rank to 64.

[xxoospring]

run lora inference will report error when i modify the lora rank to 64.

lora_ft_webui_infer.py#Line269, "lora_selection" should be pass to the load_model function