https://github.com/erogol/tts_tf
WIP Tensorflow implementation of https://github.com/mozilla/TTS
https://github.com/erogol/tts_tf
deep-learning tacotron tensorflow tensorflow2 text-to-speech tts
Last synced: 12 months ago
JSON representation
WIP Tensorflow implementation of https://github.com/mozilla/TTS
- Host: GitHub
- URL: https://github.com/erogol/tts_tf
- Owner: erogol
- Created: 2019-10-16T08:20:04.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2020-04-11T10:00:55.000Z (about 6 years ago)
- Last Synced: 2025-05-06T22:09:03.035Z (about 1 year ago)
- Topics: deep-learning, tacotron, tensorflow, tensorflow2, text-to-speech, tts
- Language: Python
- Homepage:
- Size: 101 KB
- Stars: 15
- Watchers: 5
- Forks: 2
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
Tensorflow implementation of https://github.com/mozilla/TTS
### TODO
- [ ] Catching up with TTS pytorch
- [x] Replace pytorch layers with TF
- [x] Basic eager mode training
- 5x times slower than pytorch (if you see any easy optimizations, please let me know.)
- [ ] Optimization with tf.function
- tf.function looks so hard to use with models having different input shapes per iteration since TF retraces the graph for each input. As far as I see from the official examples, workarounds take the advantage of using eager mode. (Correct me if I am wrong.)
- [ ] TF based data loader
- [ ] Model exporting pipeline for deployment
- [ ] Vocoder adaptation
- [ ] WaveRNN
- [ ] LPCNet