Robust Universal Neural Vocoding

A PyTorch implementation of Robust Universal Neural Vocoding. Audio samples can be found here.

Quick Start

1. Ensure you have Python 3 and PyTorch 1.

2. Clone the repo:

git clone https://github.com/bshall/UniversalVocoding
cd ./UniversalVocoding

3. Install requirements:

pip install -r requirements.txt

4. Download and extract ZeroSpeech2019 TTS without the T English dataset:

wget https://download.zerospeech.com/2019/english.tgz
tar -xvzf english.tgz

5. Extract Mel spectrograms and preprocess audio:

python preprocess.py

6. Train the model:

python train.py

7. Generate:

python generate.py --checkpoint=/path/to/checkpoint.pt --wav-path=/path/to/wav.wav

Pretrained Models

Pretrained weights for the 9-bit model are available here.

Notable Differences from the Paper

1. Trained on 16kHz audio from 102 different speakers (ZeroSpeech 2019: TTS without T English dataset)
2. The model generates 9-bit mu-law audio (planning on training a 10-bit model soon)
3. Uses an embedding layer instead of one-hot encoding

Acknowlegements

• https://github.com/fatchord/WaveRNN