MLAAD

The MLAAD dataset/benchmark contains synthetic speech utterances and is used to evaluate the effectiveness of source verification systems in distinguishing between synthetic speech sources while accounting for speaker traits.

Papers using MLAAD (3)

MLAAD: The Multi-language Audio Anti-spoofing Dataset2024 · 59 citesFusion of Modulation Spectrogram and SSL with Multi-head Attention for Fake Speech Detection2025 · 1 citesDisentangling Speaker Traits for Deepfake Source Verification via Chebyshev Polynomial and Riemannian Metric Learning2026