IMaSC TTS

IMaSC Malayalam TTS is a state-of-the-art Malayalam text-to-speech model, optimized to run locally. Trained with the piper framework, it is designed to be lightweight and fast. It can be used with the Sonata neural voice plugin with the NVDA screen reader. The model is trained on the IMaSC dataset, and is part of a joint effort between ICFOSS and Anannia to improve accessability for the Malayalam-speaking visually impaired community.

Downloads

The model is distributed in the ONNX format and is available to download here.

Usage

Accessibility with NVDA

To use the model with the Sonata neural voice plugin for NVDA, follow the installation guide.

General Usage

For other applications, refer to the Piper documentation for detailed instructions on setting up and deploying the model.

Citation

If you use the IMaSC Malayalam TTS model in your work, please cite:

@misc{gopinath2022imasc,
    title={IMaSC -- ICFOSS Malayalam Speech Corpus},
    author={Deepa P Gopinath and Thennal D K and Vrinda V Nair and Swaraj K S and Sachin G},
    year={2022},
    eprint={2211.12796},
    archivePrefix={arXiv},
    primaryClass={cs.SD}
}

License

The model is released under the Apache 2.0 License.