Face manipulation
Looks for model evidence associated with altered or replaced faces.
Deepfake video detector for face and voice impersonation
Deepfakes are a narrower problem than all AI video. This check reviews face manipulation and synthetic voice separately from broader generated-video signals.
Specialist checks Data handling explained Uncertainty shown
Check this source What this detector reviews
Audio consistency
Checks whether a detectable voice has synthetic characteristics.
Whole-clip context
Reports duration and usable signals instead of scoring an unsupported modality.
Reading the result
Use the score as a signal, not proof
A detected signal means the model found a pattern associated with manipulation. Confirm it with the original file, the source account, reverse search, metadata and direct contact through a trusted channel.
Important limitation: A detector cannot establish who created a clip, why it was edited or whether a true clip is being used with a false caption.
Read the full methodologyWhen this check can help
Identity claims
Investigate clips presented as a real public figure or colleague.
Impersonation
Review a suspicious video message before acting on it.
Editorial review
Prioritize footage that needs deeper verification.
Questions people ask before checking
How do I check a suspected face swap or lip-sync deepfake?
Upload the earliest, highest-quality copy available. Review the face-manipulation and synthetic-voice results, then inspect timing and context manually and verify the claimed person through a trusted channel.
Is every AI-generated video a deepfake?
No. A deepfake normally imitates or alters a person. Fully generated scenes without an impersonated identity are AI video, but not necessarily a deepfake.
Can a deepfake detector identify who made the video?
No. It can flag detectable manipulation patterns, but it cannot establish the creator, motive or truth of the caption around the clip.