search

CorentinJ / Real-Time-Voice-Cloning

Clone a voice in 5 seconds to generate arbitrary speech in real-time
Other
52.45k stars 8.77k forks source link

Synthesizer model re-training for single speaker #1000

Open thasin96 opened 2 years ago

thasin96 commented 2 years ago

I have re-trained using my custom dataset for single speker. The model has been successfully trained by following this link https://github.com/CorentinJ/Real-Time-Voice-Cloning/issues/437#issuecomment-666099538. When I try to load my custom synthesizer model, it gives trained speaker's voice but it's completely irrelevant to the cloning message. The audio is generated from sample dataset. For ex, I have a sentence like this in my dataset. "I like playing cricket". My cloning text is "Welcome to toolbox". But it outputs "I like playing cricket". I am not sure what went wrong. Any help would be appreciable.

Korean88 commented 1 year ago

Can you please describe how you trained the model? What custom dataset did you use, what equipment do you have, how long did you train, etc? Also, do you mind sharing your models?