Fine Tuning VITS model gives robotic metallic sound

[Muhammad057]

Hello everyone, I have 145 training samples of around 3 seconds to 14 seconds each. I fine-tined the VITS model. The voice similarity is okay but it gives me a metallic noisy voice. Any ideas on what can I do to avoid the metallic output?

1. Should I increase the training dataset size?
2. Should I fine-tune yourTTS model and see the results?

Thank you.

[Harrolee]

Check out this thread
I had the same result early on in my training process. It gets better with more iterations.