Matcha-TTS/README.md

🍵 Matcha-TTS: A fast TTS architecture with conditional flow matching

Shivam Mehta, Ruibo Tu, Jonas Beskow, Éva Székely, and Gustav Eje Henter

This is the official code implementation of 🍵 Matcha-TTS.

We propose 🍵 Matcha-TTS, a new approach to non-autoregressive neural TTS, that uses conditional flow matching (similar to rectified flows) to speed up ODE-based speech synthesis. Our method:

• Is probabilistic
• Has compact memory footprint
• Sounds highly natural
• Is very fast to synthesise from

Check out our demo page. Read our arXiv preprint for more details.

Installation

1. Create an environment (suggested but optional)

conda create -n matcha_tts python=3.10 -y
conda activate matcha_tts

2. Install Matcha TTS using pip from source (in future we plan to add it to PyPI)

pip install git+https://github.com/shivammehta25/Matcha-TTS.git

3. Run CLI / gradio app / jupyter notebook

# This will download the required models
match_tts --text "<INPUT TEXT>"

or

matcha_tts_app

or open synthesis.ipynb on jupyter notebook

CLI Arguments

• To synthesise from given text, run:

match_tts --text "<INPUT TEXT>"

• To synthesise from a file, run:

match_tts --file <PATH TO FILE>

• To batch synthesise from a file, run:

match_tts --file <PATH TO FILE> --batched

Additional arguments

• Speaking rate

match_tts --text "<INPUT TEXT>" --speaking_rate 1.0

• Sampling temperature

match_tts --text "<INPUT TEXT>" --temperature 0.667

• Euler ODE solver steps

match_tts --text "<INPUT TEXT>" --steps 10

Citation information

If you find this work useful, please cite our paper:

@article{mehta2023matcha,
  title={Matcha-TTS: A fast TTS architecture with conditional flow matching},
  author={Mehta, Shivam and Tu, Ruibo and Beskow, Jonas and Sz{\'e}kely, {\'E}va and Henter, Gustav Eje},
  journal={arXiv preprint arXiv:2309.03199},
  year={2023}
}

Train with your own dataset

Let's assume we are training with LJSpeech

1. Download the dataset from here, extract it to data/LJSpeech-1.1, and prepare the filelists to point to the extracted data like the 5th point of setup in Tacotron2 repo.

2. Clone and enter this repository

git clone https://github.com/shivammehta25/Matcha-TTS.git
cd Matcha-TTS

3. Install the package from source

pip install -e .

4. Go to configs/data/ljspeech.yaml and change

train_filelist_path: data/filelists/ljs_audio_text_train_filelist.txt
valid_filelist_path: data/filelists/ljs_audio_text_val_filelist.txt

5. Generate normalisation statistics with the yaml file of dataset configuration

matcha-data-stats -i ljspeech.yaml
# Output:
#{'mel_mean': -5.53662231756592, 'mel_std': 2.1161014277038574}

Update these values in configs/data/ljspeech.yaml under data_statistics key.

data_statistics:  # Computed for ljspeech dataset
  mel_mean: -5.536622
  mel_std: 2.116101

to the paths of your train and validation filelists.

5. Run the training script

make train-ljspeech

or

python matcha/train.py experiment=ljspeech

• for a minimum memory run

python matcha/train.py experiment=ljspeech_min_memory

• for multi-gpu training, run

python matcha/train.py experiment=ljspeech trainer.devices=[0,1]

Acknowledgements

Since this code uses: Lightning-Hydra-Template, you have all the powers that comes with it.

Other source codes I would like to acknowledge:

• Coqui-TTS : For helping me figure out how to make cython binaries pip installable
• Grad-TTS: For source code of MAS
• torchdyn: Useful for trying other ODE solvers during development
• labml.ai: For RoPE implementation