itzune/antton-dataset

---
license: cc-by-4.0
language:
- eu
pretty_name: Antton Dataset
size_categories:
- 10k<n<100k
task_categories:
- text-to-speech
- automatic-speech-recognition
tags:
- audio
- TTS
- Basque
- Aholab
- Ilenia
- synthetic
- common-voice
base_model:
- itzune/antton-tts
---

# Antton Dataset (Synthetic)

This is a large-scale **synthetic speech corpus** designed for training and fine-tuning Basque Text-to-Speech (TTS) models. It consists of **99,996 audio files** synthesized from the "Antton" voice model.

This dataset was generated by **Itzune** and serves as the primary source for training the [itzune/antton-tts (Piper version)](https://huggingface.co/itzune/antton-tts) model.

## Dataset Structure

Due to the large volume of data (approx. 100,000 files), the dataset is organized in the **WebDataset** format. The audio files are bundled into `.tar` shards to optimize storage, I/O performance, and streaming.

### Files
- **data/**: Directory containing the `.tar` shards.
- **metadata.csv**: The main metadata file using `|` as a delimiter:
- `file_name`: The name of the audio file (e.g., `audio_1.wav`).
- `transcription`: The corresponding Basque text.

## Technical Specifications

- **Audio Format:** WAV (PCM)
- **Sample Rate:** 22050 Hz
- **Language:** Basque (eu)
- **Voice Profile:** Antton (Male)
- **Text Source:** [Mozilla Common Voice - Basque Sentence Collection](https://datacollective.mozillafoundation.org/datasets/cmj8u3p2v007tnxxbk5ng5qvh)
- **Generation Method:** Synthesized using VITS-based architecture.

## Usage

```python
from datasets import load_dataset

dataset = load_dataset("itzune/antton-dataset", streaming=True)
sample = next(iter(dataset["train"]))
print(f"Text: {sample['transcription']}")
```

## Credits and Licensing
### Source and Methodology

This is a synthetic dataset generated by Itzune. The synthesis process involved:

- **Text Acquisition**: Sentences were sourced from the Mozilla Common Voice project (Basque sentence collection).

- **Audio Synthesis**: The audio was produced using the aHoTTS synthesis tools and the pre-trained Antton (VITS) model developed by HiTZ Basque Center for Language Technology - Aholab Signal Processing Laboratory.

### Acknowledgments

Mozilla Common Voice: For providing the community-driven sentence collection.

HiTZ Basque Center for Language Technology - Aholab Signal Processing Laboratory: For the underlying synthesis technology and the Antton voice model.

Project ILENIA: The original Maider voice resource was developed with funding from Project ILENIA.

### License

Dataset Content (Audio & Text): Licensed under Creative Commons Attribution 4.0 International (CC BY 4.0).

Original Tools/Code: The aHoTTS tools used to generate this data are licensed under the Apache License 2.0.

## Citation
If you use this dataset, please cite the original work from HiTZ/Aholab:
> García, V., Hernáez, I., & Navas, E. (2022). Evaluation of Tacotron Based Synthesizers for Spanish and Basque. Applied Sciences, 12(3), 1686. https://doi.org/10.3390/app12031686