You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Privacy-preserving audio deepfake detection. Strips the spoken words with SpeechTokenizer, then watermarks and masks the remaining audio tokens. A transformer fills the masked tokens back in, and SafeEar's spectrogram transformer classifies the clip as real or fake.