README.md

metadata

dataset_info:
  config_name: tamil
  features:
    - name: chunked_audio_filepath
      dtype: audio
    - name: text
      dtype: string
    - name: pred_text
      dtype: string
    - name: audio_filepath
      dtype: string
    - name: start_time
      dtype: float64
    - name: duration
      dtype: float64
    - name: alignment_score
      dtype: float64
    - name: en_text
      dtype: string
    - name: en_mining_score
      dtype: float64
  splits:
    - name: train
      num_bytes: 36896712990
      num_examples: 55307
  download_size: 28170670411
  dataset_size: 36896712990
configs:
  - config_name: tamil
    data_files:
      - split: train
        path: tamil/train-*
language:
  - ta
  - hi
  - te

AI4B-IndicVoices-Curated-v0 🗣️

A high-quality curated subset of IndicVoices-ST dataset, processed and filtered for Text-to-Speech (TTS) applications. This dataset focuses on clean, well-aligned speech data with high-quality transcriptions.

Currently Available Languages

• Tamil (80 hours)

More languages (Hindi and Telugu) will be added soon!

Dataset Details

Tamil Dataset

• Size: ~80 hours
• Source: Curated from IndicVoices-ST
• Filtering Criteria:
  • Alignment score > 0.9
  • Mining score > 0.5
  • Duration between 1-120 seconds
  • Clear speech without background noise
  • Well-aligned transcriptions

Dataset Structure

{
    'chunked_audio_filepath': dict,  # Actual Audio of type datasets.Audio
    'pred_text': str,                # Predicted Tamil text (has no use to TTS)
    'text': str,                     # Ground truth Tamil text
    'audio_filepath': str,           # Path to audio file
    'duration': float,               # Duration in seconds
    'alignment_score': float,
    'en_mining_score': float
}

Use Cases

• Training Text-to-Speech (TTS) models
• Speech research for Indic language
• Prosody and pronunciation studies
• Benchmarking speech synthesis systems

Citation

If you use this dataset, please cite both this repository and the original IndicVoices-ST dataset:

@article{jain2024bhasaanuvaad,
  title   = {BhasaAnuvaad: A Speech Translation Dataset for 14 Indian Languages},
  author  = {Sparsh Jain and Ashwin Sankar and Devilal Choudhary and Dhairya Suman and Nikhil Narasimhan and Mohammed Safi Ur Rahman Khan and Anoop Kunchukuttan and Mitesh M Khapra and Raj Dabre},
  year    = {2024},
  journal = {arXiv preprint arXiv: 2411.04699}
}

@misc{AI4B-IndicVoices-Curated-2024,
  title={AI4B-IndicVoices-Curated: High-Quality Speech Dataset for Indian Languages},
  author={Abhinand},
  year={2024},
  publisher={HuggingFace},
  howpublished={\url{https://huggingface.co/datasets/abhinand/AI4B-IndicVoices-Curated-v0}}
}

License

This dataset follows the same license as the original IndicVoices-ST dataset.

Acknowledgments

• AI4Bharat for the original IndicVoices-ST dataset
• All contributors to the original dataset collection

Updates

• January 2025: Initial release with Tamil dataset (80 hours)
• Coming soon: Hindi and Telugu datasets

Contact

For questions or issues, please open an Issue or contact the maintainer - abhinand5.

---

Note: This is an ongoing project, and we're continuously working to add more languages and improve data quality.