search

NVIDIA / flowtron

Flowtron is an auto-regressive flow-based generative network for text to speech synthesis with control over speech variation and style transfer
https://nv-adlr.github.io/Flowtron
Apache License 2.0
887 stars 177 forks source link

need a usable pretrained waveglow vocoder #115

Closed JohnHerry closed 3 years ago

JohnHerry commented 3 years ago

@rafaelvalle
I noticed that the waveglow pretrained model provided in [https://github.com/NVIDIA/WaveGlow] may not be directly used with flowtron. Becasue in flowtron trainning waves should be firstly normalized with 32768 to compute mel-spectrogram, while this waveglow pretrained model not. So is there any other pretrained waveglow models that can be used in flowtron tts? Or we should change flowtron code in accordance with before training？

JohnHerry commented 3 years ago

Invalid question. the released pretrained waveglow proved to be ok!