Three Seconds Is All It Takes

A teenager in a bright bedroom records a selfie video on a phone; a glowing ribbon of soundwave flows out of the phone and off-frame into an old-fashioned microphone held by a shadowy hand at the edge.

AI voice cloning tools can generate a convincing replica of any voice from as little as three seconds of audio.^[1] That audio comes from social media videos, voicemails, TikToks, or any clip posted publicly online. The resulting voice does not just sound similar. It replicates breathing patterns, speech rhythms, and emotional inflections. In tests, the cloned voices have fooled people who have known the real person for decades.

Deepfake video works the same way. Using photos and short clips, AI can generate a face that moves, reacts, and appears to speak in real time. Fake celebrity investment videos and fake video calls with executives and family members are all now documented fraud methods.