README.md

metadata

license: cc-by-4.0
language:
  - de
tags:
  - speech
  - audio
  - tts
  - thorsten-voice
  - german
size_categories:
  - 10K<n<100K

Thorsten-Voice TV-24kHz-Neutral Dataset

This dataset is a resampled version of the "TV-2022.10-Neutral" configuration from the original Thorsten-Voice TV-44kHz-Full dataset, converted from 44.1kHz to 24kHz sampling rate.

Dataset Description

The Thorsten-Voice dataset contains German speech recordings by Thorsten Müller, suitable for text-to-speech (TTS) training and other speech synthesis tasks.

Changes from Original

• Sample Rate: Converted from 44.1kHz to 24kHz
• Format: Maintained in Parquet format for efficient loading
• Content: All other metadata and text content remains identical

Use Cases

• Text-to-Speech (TTS) model training
• German speech synthesis research
• Voice cloning applications
• Speech processing experiments

Loading the Dataset

from datasets import load_dataset

dataset = load_dataset("Thorsten-Voice/TV-24kHz-Neutral")

Citation

If you use this dataset, please cite the original Thorsten-Voice work:

@misc{thorsten-voice,
  title={Thorsten-Voice},
  author={Thorsten Müller},
  year={2022},
  url={https://github.com/thorstenMueller/Thorsten-Voice}
}

License

This dataset maintains the same CC-BY-4.0 license as the original Thorsten-Voice dataset.