projecte-aina/matxa-tts-cat-multispeaker

🍵 Matxa-TTS Catalan Multispeaker

Table of Contents

Click to expand

• Model description
• Intended uses and limitations
• How to use
• Training
• Evaluation
• Citation
• Additional information

Model Description

🍵 Matxa-TTS is based on Matcha-TTS that is an encoder-decoder architecture designed for fast acoustic modelling in TTS. The encoder part is based on a text encoder and a phoneme duration prediction that together predict averaged acoustic features. And the decoder has essentially a U-Net backbone inspired by Grad-TTS, which is based on the Transformer architecture. In the latter, by replacing 2D CNNs by 1D CNNs, a large reduction in memory consumption and fast synthesis is achieved.

Matxa-TTS is a non-autorregressive model trained with optimal-transport conditional flow matching (OT-CFM). This yields an ODE-based decoder capable of generating high output quality in fewer synthesis steps than models trained using score matching.

Intended Uses and Limitations

This model is intended to serve as an acoustic feature generator for multispeaker text-to-speech systems for the Catalan language. It has been finetuned using a Catalan phonemizer, therefore if the model is used for other languages it may will not produce intelligible samples after mapping its output into a speech waveform.

The quality of the samples can vary depending on the speaker. This may be due to the sensitivity of the model in learning specific frequencies and also due to the quality of samples for each speaker.

How to Get Started with the Model

Installation

This model has been trained using the espeak-ng open source text-to-speech software. The espeak-ng containing the Catalan phonemizer can be found here

Create a virtual environment:

python -m venv /path/to/venv

source /path/to/venv/bin/activate

For training and inferencing with Catalan Matxa-TTS you need to compile the provided espeak-ng with the Catalan phonemizer:

git clone https://github.com/projecte-aina/espeak-ng.git

export PYTHON=/path/to/env/<env_name>/bin/python
cd /path/to/espeak-ng
./autogen.sh
./configure --prefix=/path/to/espeak-ng
make
make install

pip cache purge
pip install mecab-python3
pip install unidic-lite

Clone the repository:

git clone -b dev-cat https://github.com/langtech-bsc/Matcha-TTS.git
cd Matcha-TTS

Install the package from source:

pip install -e .

For Inference

PyTorch

Speech end-to-end inference can be done together with Catalan Matxa-TTS. Both models (Catalan Matxa-TTS and alVoCat) are loaded remotely from the HF hub.

First, export the following environment variables to include the installed espeak-ng version:

export PYTHON=/path/to/your/venv/bin/python
export ESPEAK_DATA_PATH=/path/to/espeak-ng/espeak-ng-data
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/path/to/espeak-ng/lib
export PATH="/path/to/espeak-ng/bin:$PATH"

Then you can run the inference script:

cd Matcha-TTS
python3 matcha_vocos_inference.py --output_path=/output/path --text_input="Bon dia Manel, avui anem a la muntanya."

You can also modify the length scale (speech rate) and the temperature of the generated sample:

python3 matcha_vocos_inference.py --output_path=/output/path --text_input="Bon dia Manel, avui anem a la muntanya." --length_scale=0.8 --temperature=0.7

ONNX

We also release a ONNX version of the model

For Training

The entire checkpoint is also released to continue training or finetuning. See the repo instructions

Training Details

Training data

The model was trained on 2 Catalan speech datasets

Dataset	Language	Hours	Num. Speakers
Festcat	ca	22	11
OpenSLR69	ca	5	36

Training procedure

Catalan Matcha-TTS was finetuned from the English multispeaker checkpoint, which was trained with the VCTK dataset and provided by the model authors.

The embedding layer was initialized with the number of catalan speakers (47) and the original hyperparameters were kept.

Training Hyperparameters

• batch size: 32 (x2 GPUs)
• learning rate: 1e-4
• number of speakers: 47
• n_fft: 1024
• n_feats: 80
• sample_rate: 22050
• hop_length: 256
• win_length: 1024
• f_min: 0
• f_max: 8000
• data_statistics:
  • mel_mean: -6578195
  • mel_std: 2.538758
• number of samples: 13340

Evaluation

Validation values obtained from tensorboard from epoch 2399*:

• val_dur_loss_epoch: 0.38
• val_prior_loss_epoch: 0.97
• val_diff_loss_epoch: 2.195

(Note that the finetuning started from epoch 1864, as previous ones were trained with VCTK dataset)

Citation

If this code contributes to your research, please cite the work:

@misc{mehta2024matchatts,
      title={Matcha-TTS: A fast TTS architecture with conditional flow matching}, 
      author={Shivam Mehta and Ruibo Tu and Jonas Beskow and Éva Székely and Gustav Eje Henter},
      year={2024},
      eprint={2309.03199},
      archivePrefix={arXiv},
      primaryClass={eess.AS}
}

Additional Information

Author

The Language Technologies Unit from Barcelona Supercomputing Center.

Contact

For further information, please send an email to langtech@bsc.es.

Copyright

Copyright(c) 2023 by Language Technologies Unit, Barcelona Supercomputing Center.

License

Apache 2.0

Funding

This work has been promoted and financed by the Generalitat de Catalunya through the Aina project.