Skip to content

40% off every plan

See plans
facecam

How to Spot a Deepfake

We make deepfakes, so here is where ours give themselves away, with close-ups.

By Facecam team · Updated

Short answer

Look where the new face meets the original person: the jawline, beard and hairline, and the skin tone of the face against the neck and hands. Then watch the teeth, the eyes and fast head turns, which swaps handle worst. Finally, check the file's metadata for an AI label and where the clip came from. No single sign proves it, but two or three together usually do.

Step by step

  1. Look at the edge of the face

    A face swap replaces the inner face only. Follow the jawline, chin and hairline: a beard, stubble or hairline that doesn't belong to the face above it is the clearest tell. In our photo swap, the original man's white beard sits right under a smooth new face.

  2. Compare the face with the neck and hands

    The face is new; the neck, ears and hands aren't. Different skin tone, age or texture between them gives the swap away. In our video swap, the face is clearly darker than the neck right below it.

  3. Watch the teeth and the inside of the mouth

    Teeth come out blurred, merged into one strip, or change shape between frames. Our video swap shows softer, smeared teeth than the original.

  4. Ask for a head turn or a hand in front of the face

    Swaps are trained mostly on faces seen from the front. A full side profile, a fast turn or a hand passing over the face makes the swap slip, flicker or drop for a moment, especially live in a call.

  5. Look for skin that's too smooth

    Good swaps, like our live one, can be hard to catch at a glance. Up close the skin is a little waxy and evenly lit, with less pore detail than the hair and clothes around it.

  6. Check the file's metadata

    Many AI tools label their output. Look for Content Credentials (C2PA) or an IPTC “Digital Source Type” of trained algorithmic media in the file's details.

  7. Check where it came from

    Find the earliest copy and who posted it. A clip that only exists as a re-upload, with no original source, deserves doubt whatever it looks like.

Live Face Swap demo
Be anyone on your callsLive Face Swap

From our test

  • Close-up of a swapped face: smooth skin and lips above a thick white beard from the original photo
    Photo swap, 200% zoom: the new face ends at the jaw, the original's beard carries on below it.
  • Close-up of a video frame where the swapped face is darker than the neck below it and the teeth are smeared
    Video swap, 200% zoom: a step in skin tone at the jaw, and softer teeth.
  • Close-up of a live face swap frame: a convincing smile with slightly smooth skin and the original's stubble on the jaw
    Live swap, 200% zoom: hard to catch, but the skin is smoother than the hair and stubble around it.

If it doesn't work

ProblemFix
It looks real and none of the tells showSome swaps are very good on a still frame. Watch the clip at full size and slowed down, and check the source and metadata instead.
A detection website says it's realAutomated detectors miss new methods and get fooled by compression. Treat their answer as one hint, not proof.
It's a live video callAsk the person to turn their head fully to the side and pass a hand in front of their face. Live swaps struggle with both.

Questions

What is the easiest way to spot a deepfake?

Look where the face meets the rest of the person: the jaw, beard, hairline and neck. A face swap changes only the inner face, so mismatches show up there first.

Can you spot a deepfake on a video call?

Often, yes. Ask the person to turn fully sideways or wave a hand in front of their face: live swaps flicker or slip on side profiles and when the face is covered.

Do deepfake detectors work?

Partly. They catch many older or low-quality fakes, but miss new methods and heavily compressed clips. Use them alongside your own checks and the source.

How do deepfake detectors work?

Most are AI models trained on large sets of real and fake videos to find the traces generators leave, such as blending edges, odd frequencies in the image and mismatched lighting. Others check for provenance labels like C2PA.

Live Face Swap

Be anyone on your calls

Start