search

modelscope / FunCodec

FunCodec is a research-oriented toolkit for audio quantization and downstream applications, such as text-to-speech synthesis, music generation et.al.
https://funcodec.github.io/
MIT License
370 stars 30 forks source link

funcodec finetune #24

Closed HaiFengZeng closed 8 months ago

HaiFengZeng commented 9 months ago

how can we adapt training the codec for smaller dataset?

ZhihaoDU commented 9 months ago

how can we adapt training the codec for smaller dataset?

You can refer the recipe egs/LibriTTS/codec/run.sh to prepare your own data and finetune the model on it. You can also use the pre-trained parameters as a initialization with --init_param xxx/xxx/model.pth