neonbjb / tortoise-tts

A multi-voice TTS system trained with an emphasis on quality
Apache License 2.0
12.82k stars 1.78k forks source link

finetuning the HiFi-GAN for the fastapi #660

Open jazza420 opened 10 months ago

jazza420 commented 10 months ago

Hey I was just curious whether a finetuned HiFi-GAN on a specific voice and finetuned AR had potential and worth trying? I am too stupid to figure it out myself but maybe someone could provide a script for it. The fast api produces decent results at very fast speeds however I feel it could be even better if the HiFi-GAN was finetuned.

Liujingxiu23 commented 6 months ago

Hi, how to train the hifigan model in fast-api mode. so the inference can done like latent -> hifigan -> wave?