https://github.com/josefalbers/e2tts-mlx
Embarrassingly Easy Fully Non-Autoregressive Zero-Shot TTS (E2 TTS) in MLX
https://github.com/josefalbers/e2tts-mlx
e2-tts flow-matching mlx multistream text-to-speech transformer tts
Last synced: 3 months ago
JSON representation
Embarrassingly Easy Fully Non-Autoregressive Zero-Shot TTS (E2 TTS) in MLX
- Host: GitHub
- URL: https://github.com/josefalbers/e2tts-mlx
- Owner: JosefAlbers
- License: apache-2.0
- Created: 2024-10-06T04:06:15.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2024-10-15T12:21:33.000Z (about 1 year ago)
- Last Synced: 2025-07-01T11:07:58.956Z (4 months ago)
- Topics: e2-tts, flow-matching, mlx, multistream, text-to-speech, transformer, tts
- Language: Python
- Homepage:
- Size: 2.87 MB
- Stars: 27
- Watchers: 2
- Forks: 4
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# e2tts-mlx: Embarrassingly Easy Fully Non-Autoregressive Zero-Shot TTS in MLX
A lightweight implementation of [Embarrassingly Easy Fully Non-Autoregressive Zero-Shot TTS](https://arxiv.org/abs/2406.18009) model using MLX, with minimal dependencies and efficient computation on Apple Silicon.
## Quick Start
### Install
```zsh
# Quick install (note: PyPI version may not always be up to date)
pip install e2tts-mlx
# For the latest version, you can install directly from the repository:
# git clone https://github.com/JosefAlbers/e2tts-mlx.git
# cd e2tts-mlx
# pip install -e .
```
### Usage
To use a pre-trained model for text-to-speech:
```zsh
e2tts 'We must achieve our own salvation.'
```
This will write `tts_0.wav` to the current directory, which you can then play.
https://github.com/user-attachments/assets/89265113-0785-42f1-b867-c38c886acbef
To train a new model with default settings:
```zsh
e2tts
```
To train with custom options:
```zsh
e2tts --batch_size=16 --n_epoch=100 --lr=1e-4 --depth=8 --n_ode=32
```
Select training options:
- `--batch_size`: Set the batch size (default: 32)
- `--n_epoch`: Set the number of epochs (default: 10)
- `--lr`: Set the learning rate (default: 2e-4)
- `--depth`: Set the model depth (default: 8)
- `--n_ode`: Set the number of steps for sampling (default: 1)
- `--more_ds` parameter: Implements [two-set training](https://arxiv.org/pdf/2410.07041) (default: 'JosefAlbers/lj-speech')
## Acknowledgements
Special thanks to [lucidrains](https://github.com/lucidrains/e2-tts-pytorch)' fantastic code that inspired this project, and to [lucasnewman](https://github.com/lucasnewman/vocos-mlx)'s the Vocos implementation that made this possible.
## License
Apache License 2.0