search

KunZhou9646 / seq2seq-EVC

This is the implementation of our Interspeech 2021 paper: Limited data emotional voice conversion leveraging text-to-speech: two-stage sequence-to-sequence training.
80 stars 16 forks source link

How can I change the vocoder ? #8

Closed nguyenviettuan96 closed 2 years ago

nguyenviettuan96 commented 2 years ago

Hi KunZhou, Could you give me some answers because I try to make sound voice more clear?

  1. Is that your code using Griffin-Lim vocoder?
  2. How can I change to neutral vocoder ? Thank you
KunZhou9646 commented 2 years ago

1/ This code is using Griffin-Lim vocoder; 2/ You just need to synthesize waveform from the converted mels by using your pre-trained neural vocoders.

tranducanhbk commented 2 years ago

Hi KunZhou, I got the same problem. Can u tell detail how to change Griffin-Lim vocoder to neural vocoders in pre-trained

KunZhou9646 commented 2 years ago

You just need to synthesize the waveform from the mel folder using your neural vocoder.

nguyenviettuan96 commented 2 years ago

Yes I just have read your other issues and found that your ParallelWaveGAN pretrain https://drive.google.com/file/d/1n01r7p-XALB6jcI8QGPfX4o9C7JRo6kX/view?usp=sharing . But it is necessary to have config.yaml file to run the model, it would be great if you share it.

KunZhou9646 commented 2 years ago

config.pdf I enclose my config file here (Please change .pdf to .yml)

nguyenviettuan96 commented 2 years ago

Thank you you can close this