search

vlomme / Multi-Tacotron-Voice-Cloning

Phoneme multilingual(Russian-English) voice cloning based on
https://github.com/CorentinJ/Real-Time-Voice-Cloning
Other
385 stars 96 forks source link

WaveGlow #7

Closed cnstntn-kndrtv closed 4 years ago

cnstntn-kndrtv commented 4 years ago

Is it possible to use WaveGlow? And not to train your own model but use pretrained?

vlomme commented 4 years ago

Yes. You must replace the vocoder

cnstntn-kndrtv commented 4 years ago

Can you provide some example? Maybe we can create another branch. Goal is to increase the quality of synthesised voice without long and expensive training (WaveGlow trained on 4 GPU for about 1, or 2 weeks). May be it would be helpful for you too.

vlomme commented 4 years ago

Sorry, but I don’t have the desire and time right now.