bhushnbrt/AUDETER

---
title: "Dataset for AUDETER: A Large-scale Dataset for Deepfake Audio Detection in Open Worlds"
configs:
- config_name: mls-tts-bark
data_files:
- split: dev
path: mls/tts/bark/dev/*.tar
- split: test
path: mls/tts/bark/test/*.tar
- config_name: mls-tts-chattts
data_files:
- split: dev
path: mls/tts/chattts/dev/*.tar
- split: test
path: mls/tts/chattts/test/*.tar
- config_name: mls-tts-cosyvoice
data_files:
- split: dev
path: mls/tts/cosyvoice/dev/*.tar
- split: test
path: mls/tts/cosyvoice/test/*.tar
- config_name: mls-tts-f5_tts
data_files:
- split: dev
path: mls/tts/f5_tts/dev/*.tar
- split: test
path: mls/tts/f5_tts/test/*.tar
- config_name: mls-tts-fish_speech
data_files:
- split: dev
path: mls/tts/fish_speech/dev/*.tar
- split: test
path: mls/tts/fish_speech/test/*.tar
- config_name: mls-tts-sparktts
data_files:
- split: dev
path: mls/tts/sparktts/dev/*.tar
- split: test
path: mls/tts/sparktts/test/*.tar
- config_name: mls-tts-vits
data_files:
- split: dev
path: mls/tts/vits/dev/*.tar
- split: test
path: mls/tts/vits/test/*.tar
- config_name: mls-tts-xtts
data_files:
- split: dev
path: mls/tts/xtts/dev/*.tar
- split: test
path: mls/tts/xtts/test/*.tar
- config_name: mls-tts-yourtts
data_files:
- split: dev
path: mls/tts/yourtts/dev/*.tar
- split: test
path: mls/tts/yourtts/test/*.tar
- config_name: mls-tts-zonos
data_files:
- split: dev
path: mls/tts/zonos/dev/*.tar
- split: test
path: mls/tts/zonos/test/*.tar
- config_name: mls-vocoders-bigvgan
data_files:
- split: dev
path: mls/vocoders/bigvgan/dev/*.tar
- split: test
path: mls/vocoders/bigvgan/test/*.tar
- config_name: mls-vocoders-vocos
data_files:
- split: dev
path: mls/vocoders/vocos/dev/*.tar
- split: test
path: mls/vocoders/vocos/test/*.tar
- config_name: mls-vocoders-bigvsan
data_files:
- split: dev
path: mls/vocoders/bigvsan/dev/*.tar
- split: test
path: mls/vocoders/bigvsan/test/*.tar
- config_name: mls-vocoders-univnet
data_files:
- split: dev
path: mls/vocoders/univnet/dev/*.tar
- split: test
path: mls/vocoders/univnet/test/*.tar
- config_name: mls-vocoders-hifigan
data_files:
- split: dev
path: mls/vocoders/hifigan/dev/*.tar
- split: test
path: mls/vocoders/hifigan/test/*.tar
- config_name: mls-vocoders-melgan
data_files:
- split: dev
path: mls/vocoders/melgan/dev/*.tar
- split: test
path: mls/vocoders/melgan/test/*.tar
- config_name: mls-vocoders-multi_band_melgan
data_files:
- split: dev
path: mls/vocoders/multi_band_melgan/dev/*.tar
- split: test
path: mls/vocoders/multi_band_melgan/test/*.tar
- config_name: mls-vocoders-full_band_melgan
data_files:
- split: dev
path: mls/vocoders/full_band_melgan/dev/*.tar
- split: test
path: mls/vocoders/full_band_melgan/test/*.tar
- config_name: mls-vocoders-style_melgan
data_files:
- split: dev
path: mls/vocoders/style_melgan/dev/*.tar
- split: test
path: mls/vocoders/style_melgan/test/*.tar
- config_name: mls-vocoders-parallel_wavegan
data_files:
- split: dev
path: mls/vocoders/parallel_wavegan/dev/*.tar
- split: test
path: mls/vocoders/parallel_wavegan/test/*.tar
viewer: true
---

# Dataset for AUDETER: A Large-scale Dataset for Deepfake Audio Detection in Open Worlds

## Dataset Description

We introduce AUDETER (AUdio DEepfake TEst Range), a large-scale, highly diverse deepfake audio dataset for comprehensive evaluation and robust development of generalised models for deepfake audio detection. It consists of over 4,500 hours of synthetic audio generated by 11 recent TTS models and 10 vocoders with a broad range of TTS/vocoder patterns, totalling 3 million audio clips.

Please refer to our [GitHub repository](https://github.com/mike-qz-wang/AUDETER) for more information.

## Dataset Structure

The dataset is organized into four main corpus collections:

```
├── common_voice/ # Common Voice corpus
│ ├── tts/ # TTS models
│ │ ├── bark/
│ │ ├── chattts/
│ │ ├── cosyvoice/
│ │ ├── f5_tts/
│ │ ├── fish_speech/
│ │ ├── openvoice/
│ │ ├── sparktts/
│ │ ├── vits/
│ │ ├── xtts/
│ │ ├── yourtts/
│ │ └── zonos/
│ └── vocoders/ # Vocoders
│ ├── bigvgan/
│ ├── bigvsan/
│ ├── full_band_melgan/
│ ├── hifigan/
│ ├── melgan/
│ ├── multi_band_melgan/
│ ├── parallel_wavegan/
│ ├── style_melgan/
│ ├── univnet/
│ └── vocos/
├── in-the-wild/ # In-the-wild corpus
│ ├── tts/ # TTS models
│ └── vocoders/ # Vocoders
├── mls/ # Multilingual LibriSpeech corpus
│ ├── tts/ # TTS models
│ └── vocoders/ # Vocoders
└── peoples_speech/ # People's Speech corpus
├── tts/ # TTS models
└── vocoders/ # Vocoders
```