MLAAD: The Multi-Language Audio Anti-Spoofing Dataset

We present the MLAAD dataset, which is a multi-language dataset for the task of audio anti-spoofing. This dataset has been created using a diverse set of text-to-speech (TTS) models, and is designed to evaluate the out-of-domain generalization of anti-spoofing systems, both with respect to new languages, as well as new TTS models. Specifically, MLAAD comprises:

• 378.0 hours of synthetic voice,
• in 38 different languages,
• created with 82 TTS models, comprising 33 different architectures,
• which are trained on 14 different source datasets.

The dataset is supposed to be used in conjunction with the M-AILABS dataset. MLAAD provides only the synthetic audio, while M-AILABS provides the real audio.

Downloads:

• Dataset
• Paper
• Apache License, Version 2.0

How to cite:

@article{muller2024mlaad,
  title={MLAAD: The Multi-Language Audio Anti-Spoofing Dataset},
  author={M{\"u}ller, Nicolas M and Kawa, Piotr and Choong, Wei Herng and Casanova, Edresson and G{\"o}lge, Eren and M{\"u}ller, Thorsten and Syga, Piotr and Sperl, Philip and B{\"o}ttinger, Konstantin},
  journal={International Joint Conference on Neural Networks (IJCNN)},
  year={2024}
}