How to Spot a Deepfake Video in 60 Seconds

The eye knows more than you think

Even as deepfakes get better, real-time generation still leaves artifacts the human brain can catch — if you know where to look. Here are the six tells that survive in almost every consumer-grade deepfake we have seen in the wild.

1. Edges of the face

Watch where the face meets the hairline, jaw, or glasses. Look for a faint shimmer, a slight blur, or pixels that 'crawl' when the person turns their head. The face is generated; everything outside the face is original footage. The seam is where they meet.

2. Eyes and blinking

Early deepfakes barely blinked. Modern ones do, but unevenly. Watch for blinks that are too symmetrical, too rare, or that happen at strange moments mid-sentence. Pupils that don't dilate when lighting changes are another tell.

3. Teeth

Generative models still struggle with teeth as discrete objects. Look for a single white blur where individual teeth should be, or teeth that subtly change shape between frames.

4. Audio sync

Listen with the video muted, then watch with audio off. If the lip shapes feel slightly behind or ahead of the consonants — especially on 'p,' 'b,' and 'm' sounds — that's a sync gap.

5. Lighting consistency

The face should respond to ambient light. Shadow direction, color temperature, and reflections on glasses should match the room. Deepfaked faces often look 'too lit' or oddly flat.

6. Background interaction

Hands passing in front of the face, hair flicking forward, a microphone moving — these are where deepfakes break. The face stays static while the rest of the world acts on it.

When your eyes aren't enough

If something feels off but you can't pin it down, run the clip through one of these:

• Intel FakeCatcher — analyzes blood flow in pixels (real faces flush with the heartbeat).
• Sensity AI — commercial-grade detection used by media organizations.
• Deepware Scanner — free, browser-based, quick first pass.

No tool is 100%. Treat them as a second opinion, not a verdict.