Download Dataset

Access the TAPS dataset through our HuggingFace repository.

Access on HuggingFace

The TAPS dataset is hosted on HuggingFace for easy access and integration. Visit our HuggingFace page to download the dataset or use it directly in your projects.

Visit HuggingFace Repository

Quick Start

Using Datasets Library

from datasets import load_dataset

dataset = load_dataset("your-org/TAPS")

Manual Download

You can also download the dataset directly from the HuggingFace repository page:

  1. Visit the HuggingFace repository link above
  2. Navigate to the "Files and Versions" tab
  3. Download the desired dataset files

Dataset Structure

Training Set

  • • 50 speakers (25M/25F)
  • • 5,000 utterance pairs
  • • 12.7 hours of audio

Evaluation Set

  • • 10 speakers (5M/5F)
  • • 1,000 utterance pairs
  • • 2.6 hours of audio

Terms of Use

By using this dataset, you agree to:

  • Cite our paper in any resulting publications
  • Use the dataset only for research purposes
  • Not redistribute the dataset without permission
  • Comply with the terms specified in our license
View Full License

Need Help?

If you have questions about downloading or using the dataset, please reach out to us:

Open an issue on GitHub Email us