Plachtaa / seed-vc

State-of-the-Art zero-shot voice conversion & singing voice conversion with in context learning
GNU General Public License v3.0
647 stars 73 forks source link

How to fix the accent issue for languages containing the /r/ sound #41

Open Selectorrr opened 2 weeks ago

Plachtaa commented 2 weeks ago

Please have a more detailed description of your issue, and the samples of failure cases

Selectorrr commented 2 weeks ago

Pay attention to the very sound of the very first letter. playground.zip

Selectorrr commented 2 weeks ago

https://github.com/user-attachments/assets/49b7bf64-28e0-4cee-82c1-373188138647

Plachtaa commented 2 weeks ago

Thanks for your clear demonstration of the issue, I highly agree with your discovery. The reason could be follows:

  1. Training dataset does not contain language of the source
  2. Reference audio is not using the same language as the source

We will release training code in the near future to allow training or finetuning on custom data