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.