https://github.com/MahtaFetrat/ManaTTS-Persian-Speech-Dataset
ManaTTS is the largest open Persian speech dataset with 86+ hours of transcribed audio. Includes data collection pipeline and tools. Suitable for Persian text-to-speech models.
https://github.com/MahtaFetrat/ManaTTS-Persian-Speech-Dataset
data-collection data-preprocessing dataset-preparation forced-alignment mana-tts persian persian-speech speech-corpus speech-data-collection speech-dataset speech-processing speech-synthesis text-to-speech text-to-speech-dataset tts tts-dataset
Last synced: 4 months ago
JSON representation
ManaTTS is the largest open Persian speech dataset with 86+ hours of transcribed audio. Includes data collection pipeline and tools. Suitable for Persian text-to-speech models.
- Host: GitHub
- URL: https://github.com/MahtaFetrat/ManaTTS-Persian-Speech-Dataset
- Owner: MahtaFetrat
- License: mit
- Created: 2024-06-03T20:58:07.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2024-09-13T05:41:16.000Z (9 months ago)
- Last Synced: 2025-02-18T00:51:27.311Z (4 months ago)
- Topics: data-collection, data-preprocessing, dataset-preparation, forced-alignment, mana-tts, persian, persian-speech, speech-corpus, speech-data-collection, speech-dataset, speech-processing, speech-synthesis, text-to-speech, text-to-speech-dataset, tts, tts-dataset
- Language: Jupyter Notebook
- Homepage:
- Size: 16.4 MB
- Stars: 17
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# ManaTTS-Persian-Speech-Dataset
ManaTTS is the largest publicly accessible single-speaker Persian corpus, comprising over 100 hours of audio with a sampling rate of 44.1 kHz. It is released under the open CC-0 license, enabling educational and commercial use. This dataset is a comprehensive speech dataset for the Persian language, collected from the [Nasl-e-Mana](https://naslemana.com/) magazine. It includes a wide range of topics and domains, making it suitable for training high-quality text-to-speech models. The dataset is accompanied by a fully transparent, open-source pipeline for data collection and processing, including tools for audio segmentation and forced alignment.
## Dataset
The ManaTTS dataset can be downloaded from [this link](https://huggingface.co/datasets/MahtaFetrat/Mana-TTS). You can access a smaller, random sample of this dataset in the [sampled data directory](sample_data). These samples were selected to reflect the same distribution of match qualities as the complete dataset. For more details on match qualities, please refer to [the paper](https://arxiv.org/abs/2409.07259).
## Raw Data Crawling
The raw data for this dataset was crawled from the Nasl-e-Mana magazine website. The crawling script used for this purpose is also provided in this repository and on Google Colab in [this link](https://colab.research.google.com/drive/1_E5KYAwuCr9B8k6EPYjVErsx-7rrr8Vl?usp=sharing).
## Processing Pipeline
The following figure illustrates the overall processing pipeline used to create the ManaTTS dataset, including the steps for preproces
This pipeline is available as a Jupyter Notebook included in this repository. You can also run the notebook on Google Colab using [this link](https://colab.research.google.com/drive/1fWTy4IH2tSuOLrLSD8E8LMaUlI_Gnf-e?usp=sharing).
To run the pipeline, follow these steps:
1. Set up the required environment (details in the notebook)
2. Place the raw audio and text files in a directory named `raw`
3. Execute the cells in the notebook sequentially
## Trained TTS Model
A text-to-speech (TTS) model has been trained on the ManaTTS dataset. The code for training the model, as well as some output samples, are available in [this repository](https://github.com/MahtaFetrat/Persian-MultiSpeaker-Tacotron2). The model weights and inference instructions can be found in [this repository](https://huggingface.co/MahtaFetrat/Persian-Tacotron2-on-ManaTTS).
## Contributing
Contributions to this project are welcome! If you encounter any issues or have suggestions for improvements, please open an issue or submit a pull request.
## License
The ManaTTS dataset is released under the CC-0 1.0 license, while the processing pipeline is licensed under the MIT license.
## Important Notice on Ethical Use of ManaTTS Dataset
The ManaTTS dataset is provided exclusively for research and development purposes. We emphasize the critical importance of ethical conduct in utilizing this dataset. Please refrain from any misuse, including but not limited to voice impersonation, identity theft, or fraudulent activities.
By accessing and using the ManaTTS dataset, you are obligated to uphold the highest standards of integrity and respect for user privacy. Any violation of these principles may have severe legal and ethical consequences.
For any inquiries or clarifications regarding the use of this dataset, please reach out to us. Your cooperation in ensuring responsible use of this dataset is greatly appreciated.
## Acknowledgment
We would like to express our sincere gratitude to [Nasl-e-Mana](https://naslemana.com/), the monthly magazine of the blind community of Iran, for their generosity. Their commitment to openness and collaboration has been instrumental in advancing research and development in speech synthesis. We are especially thankful for their choice to release the data under the Creative Commons CC-0 license, allowing for unrestricted use and distribution.
## Collaboration and Community Impact
We encourage researchers, developers, and the broader community to utilize the resources provided in this project, particularly in the development of high-quality screen readers and other assistive technologies to support the Iranian blind community. By fostering open-source collaboration, we aim to drive innovation and improve accessibility for all.
## Citation
If you use this dataset or the processing pipeline in your work, please cite the following paper:
```bash
@article{fetrat2024manatts,
title={ManaTTS Persian: a recipe for creating TTS datasets for lower resource languages},
author={Mahta Fetrat Qharabagh and Zahra Dehghanian and Hamid R. Rabiee},
journal={arXiv preprint arXiv:2409.07259},
year={2024},
}
```