NVIDIA / tacotron2

Tacotron 2 - PyTorch implementation with faster-than-realtime inference
BSD 3-Clause "New" or "Revised" License
5.06k stars 1.38k forks source link

How does pretrained model works for any language. #568

Open A-d-DASARE opened 2 years ago

A-d-DASARE commented 2 years ago

Can anyone pls help me understand how the check points of LJ Speech which is essentially English language, can be used to fine tune any language. I used it for Kannada which is no where close to English yet could get better result. Someone pls throw some light on this. Thanks.

A-d-DASARE commented 2 years ago

No answers yet!

yuyushang commented 2 years ago

It may could be explained by that the pretrained model has learned how to get human-like voice from a phoneme sequence.

meiming24 commented 2 years ago

I think it only work with English

sunbeibei-hub commented 8 months ago

Does it apply to Japanese?