Open jinqiupeter opened 4 months ago
Hi,
I conducted experiments using the English benchmark dataset, and if possible, I will also add the Chinese version.
Thanks.
That would be great! Recently there is a new Chinese dataset https://huggingface.co/datasets/Wenetspeech4TTS/WenetSpeech4TTS that has quite good variety and prosody comparing to AISHELL-3.
Hi, this is a great repo! I'm really curious about the progress achieved up to now. Or whether multi-language adaptation can be accomplished through SFT?
@winddori2002
It seems there are no Chinese samples from the demo page, would be good to have multi-language support!