Training multiband melgan on predicted mel

TensorSpeech / TensorFlowTTS

:stuck_out_tongue_closed_eyes: TensorFlowTTS: Real-Time State-of-the-art Speech Synthesis for Tensorflow 2 (supported including English, French, Korean, Chinese, German and Easy to adapt for other languages)

https://tensorspeech.github.io/TensorFlowTTS/

Apache License 2.0

3.81k stars 811 forks source link

Training multiband melgan on predicted mel #755

Closed cnamuangtoun closed 2 years ago

cnamuangtoun commented 2 years ago

Hi, I'm just curious if training the multiband melgan with predicted mel spectogram from fastspeech2 would imrpove the quality of the waveform output from the vocoder instead of using the ground truth mel spectogram straight from the dataset

xianglan010 commented 2 years ago

I try to use the mel spec generated from pertained tacotorn2 to train the MB-melgan, but the loss not convergence, have you tried it and how about ur result

stale[bot] commented 2 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs.