Audio samples of multi-band MelGAN vocoder
The vocoder was trained for 500K iterations on the
LJSpeech dataset
Source code:
https://github.com/che-roman/mb-melgan
Use headphones to compare the original and predicted sound 🎧
Losses
Spectrograms of the sample
original
predicted
Audio samples
original
predicted
Copyright © 2023 Chudnov Roman. All rights reserved.
