dataset_info: features: - name: text dtype: string - name: audio dtype: audio - name: duration dtype: float64 splits: - name: train num_bytes: 20860321574.68 num_examples: 104520 download_size: 20907025567 dataset_size: 20860321574.68 configs:
- config_name: default
data_files:
- split: train path: data/train-*
dataset_description: >
This is a high-quality Text-to-Speech (TTS) dataset for the Urdu language.
It contains 104,520 audio-text pairs, each sampled at 22,500 Hz mono, and is ideal for building neural TTS systems.
Highlights:
- Clean and phonetically rich Urdu text
- Studio-quality audio recordings
- Mono channel at 22.5 kHz sampling rate
- Includes duration metadata for each clip
- Perfect for models like Tacotron, FastSpeech, Glow-TTS, and VITS
Format:
- text: Urdu transcription
- audio: High-quality mono audio (22.5 kHz)
- duration: Length of the audio clip in seconds
License:
This dataset is released for research and educational purposes only.
Commercial use is strictly prohibited.
Please refer to the accompanying LICENSE file for full terms and conditions.
Citation:
If you use this dataset in your research, please cite the dataset creators appropriately.