Training HiFIGAN for XTTSv2 #3659
Unanswered
JahidBasher
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
-
As per my understanding, hifigan_decoder use gpt_latent and speaker_embedding to convert mel_spec to audio signal. But hifigan was possibly trained on audio segment instead of full audio. This necessitates having audio segment to align with gpt_latent for loss calculation. How can I align it? I gave coded the following snippet for hifigan dataloader. Is it the correct way to do it?
Beta Was this translation helpful? Give feedback.
All reactions