AI Video Detector
Check whether a video was filmed or generated. We pull frames from the clip and run each one through the full forensic stack, then look for something a single image can never show you — whether the world stays consistent from one frame to the next.
What betrays a generated video
Generated video has to solve a problem still images do not: staying consistent over time. That is where it fails first, and it is why we analyse a clip as a sequence rather than as a pile of separate pictures.
Objects that don't persist
A pattern on a shirt, a logo, a background sign — they drift or redraw themselves between frames. Real objects stay exactly what they were.
Physics that almost works
Hair, cloth and water move plausibly for a second, then behave in a way no real material would. Contact between objects is where it shows most.
Faces that reset
Across a longer clip a generated face subtly changes proportion — the distance between the eyes, the shape of the jaw — because each frame is reconstructed rather than filmed.
Impossible camera motion
Real footage carries the signature of a physical camera: micro-shake, rolling shutter, focus breathing. Generated motion is unnaturally smooth or slides in ways a lens cannot.
Per-frame generation traces
Every extracted frame still carries the same statistical fingerprints we look for in stills — frequency profile, noise, texture coherence.
Missing capture metadata
Real footage records a codec, a device and a capture profile. Generated files usually carry only an export signature, or nothing at all.
How the analysis works
Upload the original file
Full resolution, straight from the source. Screenshots and re-saved copies destroy most of the evidence we measure.
12 independent methods run in parallel
Three neural networks look at the image as a whole. Nine forensic methods examine it as a signal: frequency (FFT) analysis, noise pattern, pixel-level statistics, EXIF metadata, Error Level Analysis, texture coherence, colour distribution, Benford's Law and edge detection.
You see the breakdown, not just a number
Every method reports separately, so you can tell whether the verdict rests on one weak signal or on nine that agree.
How much each method counts
Most detectors give you one number from one model. We run twelve independent analyzers and weight them — three neural networks carry the majority, but nine forensic methods examine the file as a signal and can overrule them. Here is the exact weighting we use in production:
| Method | Weight in final score |
|---|---|
| Neural network #2 | 25% |
| Neural network #1 | 15% |
| Neural network #3 | 15% |
| EXIF metadata | 10% |
| Frequency (FFT) | 6% |
| Benford's Law | 6% |
| Noise pattern | 5% |
| Error Level Analysis | 4% |
| Texture coherence | 4% |
| Edge detection | 4% |
| Pixel statistics | 3% |
| Colour distribution | 3% |
What people actually upload
These are our own numbers, not an industry estimate — every check run through FakeSec since launch. The split is closer than most people expect: roughly four in ten uploads really do turn out to be generated.
Who checks video
- Newsrooms — A clip going viral before anyone can verify where it came from.
- Moderators — Platforms now require disclosure of synthetic media.
- Buyers of stock footage — Paying for filmed material and receiving generated.
- Anyone in a dispute — Video used as evidence of something that never happened.
Accuracy — and where it breaks down
On our internal benchmark of 240,000 images (120,000 AI-generated, 120,000 real) the combined score reaches up to 98% on unedited, full-resolution files. That number drops, sometimes sharply, in four situations — and you should know them before trusting any result:
- Heavy compression. Social platforms re-encode everything. A photo pulled from Instagram has already lost much of its noise signature.
- Screenshots. A screenshot is a new image of an image. Original metadata and compression history are gone.
- Stylised content. Flat colour and clean lines remove the texture and noise evidence forensic methods depend on.
- Post-processing. Added grain, filters, upscaling and manual retouching all mask generation traces.
No detector on the market is 100% accurate — ours included. Treat a high score as strong evidence, not as proof. When a decision matters, look at which of the 12 methods agreed.
Video is harder than stills, and we will not pretend otherwise. Platform re-encoding destroys much of the forensic signal, so a clip downloaded from social media gives a weaker result than the original file. Short clips also give fewer frames to compare, which is why confidence rises with clip length.
Frequently asked questions
Related tools
Check your first image now
Create a free account and get 3 checks. Full breakdown across all 12 methods, no card required.