we have created a multilingual audio deepfake dataset, named MADD. MADD's sources come from Common Voice and GigaSpeech2. It leverages deep speech synthesis and voice conversion technologies across 6 languages, resulting in a dataset that includes 129,990 synthesized utterances and spans approximately 155.66 hours of audio.
