---
dataset_info:
  features:
  - name: file_name
    dtype:
      audio:
        sampling_rate: 22050
  - name: text
    dtype: string
  splits:
  - name: train
    num_bytes: 13446906647.04
    num_examples: 20444
  download_size: 14234302110
  dataset_size: 13446906647.04
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
language:
- ur
license: mit
task_categories:
- text-to-audio
- text-to-speech
size_categories:
- 10K<n<100K
---
# UAT: Urdu Audio Transcriptions

UAT is a dataset featuring high quality audios and their transcriptions, sourced from Urdu audiobooks. This dataset is designed to support both TTS and ASR research in the Urdu language.

## Data Format

Each entry in the dataset includes:
- **file_name**: A unique identifier for the audio file.
- **text**: The corresponding text transcription.

## Usage

This dataset is ideal for:
- Training and evaluating TTS systems.
- Developing ASR models.
- Exploring linguistic and acoustic features of the Urdu language.

## Acknowledgments

UAT is distributed under the MIT license. When using this dataset, please acknowledge the creators by citing it as follows:

```bibtex
@dataset{uat,
  title       = {UAT: Urdu Audio Transcriptions},
  author      = {Mahwiz Khalil},
  year        = {2025},
  publisher   = {Hugging Face},
  howpublished= {\url{https://huggingface.co/datasets/mahwizzzz/UAT}},
}
```
For any inquiries or further information, please contact: khalilmahwiz@gmail.com