search

jik876 / hifi-gan

HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis
MIT License
1.92k stars 506 forks source link

Fine-Tuning on multispeaker multilingual noisy dataset #108

Open bhavuk0909 opened 2 years ago

bhavuk0909 commented 2 years ago

Hi, Thanks for making your work publicly available.

I want to ask a few things:

  1. Can I get average quality audios from the noisy audio dataset?
  2. Even in a single audio file (in my dataset) there is more than one speaker speaking more than one language (English + Hindi) in a noisy environment, so will it be able to produce audios even from this type of dataset.
  3. Which one of your pre-trained models can I use for this task?

Any help or suggestions is highly appreciated. Thank you!