Retrieval-based-Voice-Conversion-WebUI ContentVec settings

Hi, I am able to train my own hubert_base with ContentVec project.

But the output hubert_base doesn't work with RVC. I believe that it is beacause some settings or configuration during training with ContentVec. Probably there in this file which is basically trigger for training. https://github.com/auspicious3000/contentvec/blob/main/run_pretrain_single.sh

@RVC-Boss Could developers please share this inforamtion? What settings do I have to set when training hubert_base for RVC? Thank you very much.

Jun 14 '24 17:06 Lukysoon

You can print the keys of the pytorch weight dict hubert_base.pth and align them.

Jun 24 '24 08:06 RVC-Boss

Hello, @Lukysoon I would like to improve the pronunciation of the RVC in other languages, and after having tested different pretrain models, I came to the conclusion that it would be necessary to train contenvec with the desired language.

However, I can't understand the instructions in the ContentVec repository, is there any tutorial or resources where I can get more enlightening information about the training and the necessary files, or how to generate them?

Thank you very much in advance.

Aug 27 '24 16:08 Mixomo