Tunisian Dialect Text-To-Speech

This project project was developed with the purpose of bridging the gap between high-resource and low-resource languages. Our main focus in creating this dataset is to work on a mono-speaker Tunisian dialect Text-to-Speech model.

This repository contains all the scraping and pre-processing scripts for Tunisian TTS dataset. It contains also the two approaches described in the paper (from scratch and transfer learning).

You can test our models in this Huggingface space. Or you can train your models following the scripts provided in this repo.

Steps :

• First of all, you have to install the requirements :
  pip install -r requirements.txt

• Then, You have to create the dataset following the scripts provided in dataset_recreation\

• Finally, you have to train the model either from scratch or using transfer learning technique as described in espnet_experiments.

If you use TunArTTS in your research, please cite it using the following BibTeX entry:

@inproceedings{TunArTTS,
    title = "TunArTTS: Tunisian Arabic Text-To-Speech Corpus",
    author = "Imen, Laouirine and Rami Kammoun  and Fethi BOUGARES",
    booktitle = "THE 2024 JOINT INTERNATIONAL CONFERENCE ON COMPUTATIONAL LINGUISTICS, LANGUAGE RESOURCES AND EVALUATION",
    month = mai,
    year = "2024",
    address = "TORINO, ITALIA",
    publisher = "European Language Resources Association",
    }