LJSpeech Traine Tacotron 2 + universal MelGan Vocoder [Multiband & FullBand MelGan] Problem #1016
Unanswered
Wayne-wonderai
asked this question in
General Q&A
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
When we attached LJSpeech pretrained Tacotron 2, to Melgan vocoders (universal ones) it doesn't seem to sound right.
However when looking at the samples, this configuration should work? We are using the pretrained models available on the repo.
We have attached what it sounds like.
If tacotron 2 and melgan arent compatible, how is it that papers have samples with this configuration?
https://github.com/rosinality/melgan-pytorch
Sample.zip
Beta Was this translation helpful? Give feedback.
All reactions