how to set the mbmelgan's parameters

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

3.8k stars 810 forks source link

how to set the mbmelgan's parameters #710

Closed hertz-pj closed 2 years ago

hertz-pj commented 2 years ago

if i want to train by the wav with "sampling_rate=16k, hop_size=200，win_size=800"。 The mbmelgan's parameters i set upsampling_scale=[2,5,5]. Do i need to set downsample_scale=[2,5,5]？ Or keep [4,4,4] unchanged?

ZDisket commented 2 years ago

@hertz-pj I've trained mb-melgan on many different sampling rates and got good results without changing the downsample scales, so I believe you should keep it as-is. But if you do have the spare time and compute feel free to experiment; who knows, you might just discover something!

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.