search

r9y9 / deepvoice3_pytorch

PyTorch implementation of convolutional neural networks-based text-to-speech synthesis models
https://r9y9.github.io/deepvoice3_pytorch/
Other
1.97k stars 485 forks source link

Training issues #200

Closed ilyalasy closed 2 years ago

ilyalasy commented 4 years ago

Hello, I'm trying to train deepvoice on russian single speaker dataset. Hyperparameters are the same as nyanko_ljspeech. After 660k steps predicted speech is terribly noisy and impossible to understand. I guess I have to tweak some hyperparameters but I dont know what exactly. Alignments of the 660k steps prediction: step000660000_text5_single_alignment