waveglow-cli 0.0.1

Command-line interface (CLI) to train WaveGlow using .wav files.

waveglow-cli

Command-line interface (CLI) to train WaveGlow using .wav files.

Features

• train/synthesize on CPU or GPU
• download pre-trained models by Nvidia

Installation

pip install waveglow-cli --user

Usage

usage: waveglow-cli [-h] [-v] {download,train,continue-train,validate,synthesize} ...

This program trains WaveGlow.

positional arguments:
  {download,train,continue-train,validate,synthesize}
                     description
    download         download pre-trained checkpoint from Nvidia
    train            start training
    continue-train   continue training
    validate         validate checkpoint(s)
    synthesize       synthesize mel-spectrograms into an audio signal

optional arguments:
  -h, --help         show this help message and exit
  -v, --version      show program's version number and exit

Pretrained Models

• LJS-v3-580000: Adapted model trained on LJ Speech dataset by Nvidia.

Audio Example

"The North Wind and the Sun were disputing which was the stronger, when a traveler came along wrapped in a warm cloak." Listen here (headphones recommended)

Dependencies

• torch
• mel-cepstral-distance>=0.0.1
• pandas
• librosa
• plotly
• scikit-image
• matplotlib
• scikit-learn
• tqdm
• wget
• gdown
• Unidecode
• Pillow
• fastdtw
• numpy
• scipy
• ordered_set>=4.1.0

Roadmap

• Outsource method to convert audio files to mel-spectrograms before training
• Improve logging
• Add more audio examples
• Adding tests

License

MIT License

Acknowledgments

Model code adapted from Nvidia.

Paper:

• Waveglow: A Flow-based Generative Network for Speech Synthesis

Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – Project-ID 416228727 – CRC 1410

Citation

If you want to cite this repo, you can use this BibTeX-entry:

@misc{tsw22,
  author = {Taubert, Stefan},
  title = {waveglow-cli},
  year = {2022},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/stefantaubert/waveglow}}
}