search

RVC-Project / Retrieval-based-Voice-Conversion-WebUI

Easily train a good VC model with voice data <= 10 mins!
MIT License
22.69k stars 3.41k forks source link

Generated audio became blurred. #292

Open skeleton66 opened 1 year ago

skeleton66 commented 1 year ago

I am training with COLAB, but the voice of the model trained yesterday and today seems noisier and blurred compared to the voice of the previously trained model. It feels like the pre-trained models are not working. The former is the model learned previously and the latter is the model learned today. https://www.axfc.net/u/4065483 Although trained on the same data set, the sound is clearly degraded in the latter. https://www.axfc.net/u/4065492 I tried learning at 40k and it was clear. I tried learning at 48k and it seemed to cause problems.

RVC-Boss commented 1 year ago

It seems that I should change the 48k config. I will check this in the future.