https://github.com/erogol/fftnet

FFTNet vocoder implementation
https://github.com/erogol/fftnet

deep-learning fftnet pytorch text2speech vocoder

Last synced: 12 months ago
JSON representation

FFTNet vocoder implementation

• Host: GitHub
• URL: https://github.com/erogol/fftnet
• Owner: erogol
• License: mpl-2.0
• Created: 2018-06-25T13:25:40.000Z (almost 8 years ago)
• Default Branch: master
• Last Pushed: 2018-09-28T17:33:52.000Z (over 7 years ago)
• Last Synced: 2025-04-30T07:36:27.924Z (about 1 year ago)
• Topics: deep-learning, fftnet, pytorch, text2speech, vocoder
• Language: Jupyter Notebook
• Size: 756 KB
• Stars: 81
• Watchers: 14
• Forks: 8
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE.txt

Awesome Lists containing this project

README

          
## Unofficial Implementation of [FFTNet vocode](http://gfx.cs.princeton.edu/pubs/Jin_2018_FAR/fftnet-jin2018.pdf) paper.

- [x] implement the model.

- [x] implement tests.

- [x] overfit on a single batch (sanity check).

- [x] linearize weights for eval time.

- [x] measure the run-time on GPU and CPU. (1 sec audio takes ~47 secs) If anyone knows additional tricks from the paper, let me know. So far I asked the authors but nobody returned. 

- [ ] train on LJSpeech spectrograms.

- [ ] distill model as in Parallel WaveNet paper.