GPT-SoVITS
GPT-SoVITS copied to clipboard
Pretrained models language
The filenames for the pretrained models include "Chinese" specifically. Would it be better for other languages to use a different version of these models? For example if i wanted to run GPT-SoVITS on english should I replace these? chinese-hubert-base ->https://huggingface.co/facebook/hubert-base-ls960 chinese-roberta-wwm-ext-large -> https://huggingface.co/FacebookAI/roberta-large
Maybe it will be better. And you should train your own base model using your proposed hubert and bert models.