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

Required features in Jan. 2024 #10

Closed ZhihaoDU closed 8 months ago

ZhihaoDU commented 11 months ago

Hi, all. I'm collecting the required features which will be considered implementing in Jan. 2024. Please let me know your concern and feel free to comment below. Thanks. To make FunCodec better!

duj12 commented 10 months ago

Hi, all. I'm collecting the required features which will be considered implementing in Jan. 2024. Please let me know your concern and feel free to comment below. Thanks. To make FunCodec better!

Support Chinese. Maybe use your 60k hour Mandarin speech data to train a LauraTTS model, and this will be a Paraformer-like TTS model in Chinese Community.

ZhihaoDU commented 10 months ago

Hi, all. I'm collecting the required features which will be considered implementing in Jan. 2024. Please let me know your concern and feel free to comment below. Thanks. To make FunCodec better!

Support Chinese. Maybe use your 60k hour Mandarin speech data to train a LauraTTS model, and this will be a Paraformer-like TTS model in Chinese Community.

Thanks for your suggestion. We are considering to train a data-scale-up model.

didadida-r commented 10 months ago

Support Chinese.