search

keonlee9420 / PortaSpeech

PyTorch Implementation of PortaSpeech: Portable and High-Quality Generative Text-to-Speech
MIT License
331 stars 36 forks source link

Is that possible train Chinese? #10

Closed luohao123 closed 2 years ago

luohao123 commented 3 years ago

Is that possible train Chinese?

luohao123 commented 3 years ago

Any plan to add Chinese support?

keonlee9420 commented 3 years ago

Hello @luohao123. Yes, since data preparation is the same as FastSpeech2. But I don't plan for that right now. It would be great if you, as a native, could adopt the Chinese dataset into this project. Maybe I can help you with some points.

luohao123 commented 3 years ago

@keonlee9420 thank u, I might start from aishell3, do you think is that possible train a tts model with multi speaker dataset?

keonlee9420 commented 3 years ago

For sure, yes. But you should do some experiments on the conditioning method in LinguisticEncoder and flows in each module. I thought of a typical speaker conditioner for the former and the technique that Glow-TTS used for the latter.

keonlee9420 commented 2 years ago

Close due to inactivity.