search

FunAudioLLM / CosyVoice

Multi-lingual large voice generation model, providing inference, training and deployment full-stack ability.
https://funaudiollm.github.io/
Apache License 2.0
5.41k stars 557 forks source link

微调 #5

Closed suzhenghang closed 3 months ago

suzhenghang commented 3 months ago

训练细节好像很少，如果要用自有数据微调，可以怎么做？

ZhihaoDU commented 3 months ago

我们在这里提供了以LibriTTS为例的微调recipe https://github.com/FunAudioLLM/CosyVoice/tree/main/examples/libritts/cosyvoice

plutols commented 3 months ago

这个是以英文微调为例，还是只支持英文的微调呢

yubinyes commented 3 months ago

请问按照这个脚本训练出来的模型，就可以支持tts和voice clone吗？跨语言inference_cross_lingual的话，中英数据集都需要吗？

aluminumbox commented 3 months ago

请问按照这个脚本训练出来的模型，就可以支持tts和voice clone吗？跨语言inference_cross_lingual的话，中英数据集都需要吗？

yes, it can, and you need both english and chinese data for cross_lingual inference