Yoruba Text-to-Speech Package

A simple Node.js package to generate speech from text using the Meta's mms-tts-yor model on HuggingFace API. It is also possible to make use of another model which might be more advance than this one.

Installation

npm install yoruba-tts

Usage

Remember to get your API token from HuggingFace and put in the .env file in the root of your project.

TTS_MODEL_URL=https://api-inference.huggingface.co/models/facebook/mms-tts-yor
HUGGINGFACE_API_KEY=

import { generateSpeech } from 'yoruba-tts'

const text = "Mo fẹ́ràn Èdè Yorùbá";
const outputPath = "./output.flac";
generateSpeech(text, outputPath);

The output will be a flac file in the location specified.

Sample Audio

Sample Audio: feran.flac

Credit

The credit for the model belongs to Vineel Pratap et al. from Meta AI.

@article{pratap2023mms,
    title={Scaling Speech Technology to 1,000+ Languages},
    author={Vineel Pratap and Andros Tjandra and Bowen Shi and Paden Tomasello and Arun Babu and Sayani Kundu and Ali Elkahky and Zhaoheng Ni and Apoorv Vyas and Maryam Fazel-Zarandi and Alexei Baevski and Yossi Adi and Xiaohui Zhang and Wei-Ning Hsu and Alexis Conneau and Michael Auli},
    journal={arXiv},
    year={2023}
}

Contact

The author can be contacted for feedback at Surajudeen Akande

License

MIT License