Hugging Face's logo

Datasets:
MahtaFetrat
/
Mana-TTS

Modalities:
Tabular
Text
Formats:
parquet
Languages:
Persian
Size:
10K - 100K
Tags:
text-to-speech
tts
speech-synthesis
persian
data-collection
data-preprocessing
DOI:
Libraries:
Datasets
Dask
License:
Dataset card Data Studio Files Community
2
Mana-TTS / README.md
MahtaFetrat's picture
MahtaFetrat
Update README.md
698f7f7 verified
preview code
|
raw
history blame contribute delete
5.62 kB
metadata
license: cc0-1.0
language:
  - fa
tags:
  - text-to-speech
  - tts
  - speech-synthesis
  - persian
  - data-collection
  - data-preprocessing
  - speech-processing
  - forced-alignment
  - speech-dataset
  - speech-corpus
  - dataset-preparation
  - persian-speech
  - tts-dataset
  - text-to-speech-dataset
  - mana-tts
  - manatts
  - speech-data-collection

ManaTTS-Persian-Speech-Dataset

ManaTTS is the largest publicly available single-speaker Persian corpus, comprising over 114 hours of high-quality audio (sampled at 44.1 kHz). Released under the permissive CC-0 license, this dataset is freely usable for both educational and commercial purposes.

Collected from Nasl-e-Mana magazine, the dataset covers a diverse range of topics, making it ideal for training robust text-to-speech (TTS) models. The release includes a fully transparent, open-source pipeline for data collection and processing, featuring tools for audio segmentation and forced alignment. For the full codebase, visit the ManaTTS GitHub repository.

Dataset Columns

Column Name Description
file_name Unique identifier for the audio file.
transcript Ground-truth text transcription of the audio chunk.
duration Duration of the audio chunk (in seconds).
match_quality Quality of alignment between the approximate transcript and ground truth (HIGH or MIDDLE). Reflects confidence in transcript accuracy (see paper for details).
hypothesis Approximate transcript used to search for the ground-truth text.
CER Character Error Rate between the hypothesis and accepted transcript.
search_type Indicates whether the transcript was matched continuously in the source text (type 1) or with gaps (type 2).
ASRs Ordered list of ASRs used until a match was found.
audio Audio file as a numerical array.
sample_rate Sampling rate of the audio file (44.1 kHz).

Usage

Python (Hugging Face)

First install the required package:

pip install datasets

Then load the data:

from datasets import load_dataset

# Load a specific partition (e.g., part 001)
dataset = load_dataset("MahtaFetrat/Mana-TTS", 
                      data_files="dataset/dataset_part_001.parquet", 
                      split="train")

# Inspect the data
print(dataset)
print(dataset[0])  # View first sample

Command Line (wget)

Download individual files directly:

# Download single file (e.g., part 001)
wget https://huggingface.co/datasets/MahtaFetrat/Mana-TTS/resolve/main/dataset/dataset_part_001.parquet

Trained TTS Model

Hugging Face

A Tacotron2-based TTS model trained on ManaTTS is available on Hugging Face. For inference and weights, visit the model repository.

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

This dataset is released under the CC-0 1.0 license.

Ethical Use Notice

The ManaTTS dataset is intended exclusively for ethical research and development. Misuse—including voice impersonation, identity theft, or fraudulent activities—is strictly prohibited. By using this dataset, you agree to uphold integrity and privacy standards. Violations may result in legal consequences.

For questions, contact the maintainers.

Acknowledgments

We extend our deepest gratitude to Nasl-e-Mana, the monthly magazine of Iran’s blind community, for their generosity in releasing this data under CC-0. Their commitment to open collaboration has been pivotal in advancing Persian speech synthesis.

Community Impact

We encourage researchers and developers to leverage this resource for assistive technologies, such as screen readers, to benefit the Iranian blind community. Open-source collaboration is key to driving accessibility innovation.

Citation

If you use ManaTTS in your work, cite our paper: 

@inproceedings{qharabagh-etal-2025-manatts,
    title = "{M}ana{TTS} {P}ersian: A Recipe for Creating {TTS} Datasets for Lower-Resource Languages",
    author = "Qharabagh, Mahta Fetrat and Dehghanian, Zahra and Rabiee, Hamid R.",
    booktitle = "Proceedings of the 2025 Conference of the North American Chapter of the Association for Computational Linguistics",
    month = apr,
    year = "2025",
    address = "Albuquerque, New Mexico",
    publisher = "Association for Computational Linguistics",
    pages = "9177--9206",
    url = "https://aclanthology.org/2025.naacl-long.464/",
}

Aditional Links