Retrieval-based-Voice-Conversion-WebUI Female to Male Voice Help

Thank you for this interesting repository! I was hoping you could give me some tips.

I've implemented the Colab code, training with around 9 minutes of one spoken male voice split into 10 second files. I've kept the standard settings (20 epochs, etc). I've also tried 40 minute audio with 100+ epochs for the heck of it.

I tried to convert a female voice using this model, and no matter how I change the transpose values, the voice isn't close to the original male voice.

Do you have any advice on how to better convert female to male voices? Or is this repository not really meant for that?

May 07 '23 18:05 mdyfong

20 epochs is not enough for almost any dataset. I train to max, or close to max, for my models

May 08 '23 09:05 kalomaze

20 epochs is not enough for almost any dataset. I train to max, or close to max, for my models

Thanks for your insight! What does max look like for you? Or do I just have to keep training (increasing epochs) until I get better results?

May 08 '23 09:05 mdyfong