Deepfake Detector
Check whether a face has been swapped, synthesised or manipulated. A deepfake fails differently from a fully generated image — the face is fake but the scene around it is real, and that seam is exactly what our forensic methods look for.
Where a deepfake gives itself away
A deepfake is a composite: a generated or borrowed face welded onto a real body in a real scene. Almost every reliable tell comes from the boundary between the two, or from the mismatch between them.
The blend boundary
Where the fake face meets real skin — hairline, jaw, ears — there is a transition zone with its own compression history. Error Level Analysis highlights it as a region that was saved a different number of times than the rest of the frame.
Two different noise floors
Every camera sensor leaves a characteristic noise pattern across the whole image. A pasted face carries a different one, or none at all. Our noise analyzer compares the face region against the background.
Lighting that disagrees
The face is lit by whatever the source footage had, the scene by its own lights. Shadows under the nose, catchlights in the eyes and the direction of highlights stop agreeing with the rest of the shot.
Teeth, ears and jewellery
Generators still handle these badly: teeth that merge into a single block, an earring present on one side only, a mismatched ear shape — ears are rarely in the training focus and are as individual as fingerprints.
Resolution mismatch
The swapped region is usually generated at a fixed size and then scaled to fit. It ends up softer or sharper than the surrounding image — measurable as a frequency (FFT) discontinuity.
Metadata that contradicts the picture
EXIF says one camera and one edit history; the pixels say a face was replaced afterwards. Missing metadata on a supposedly straight-from-camera photo is itself a signal.
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.
When you need a deepfake check
- Video calls and job interviews — Real-time face swaps are now used in remote hiring fraud — grab a frame and check it.
- Blackmail and sextortion — Fabricated compromising images are the most common abuse. Evidence that it is synthetic changes everything.
- Public figures and viral clips — A statement that never happened, spreading before anyone verifies it.
- Identity and KYC — A submitted selfie that was never taken by the person submitting it.
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.
Deepfakes are a moving target: detection is strongest on photos and individual frames, and weakest on heavily compressed video re-shared through several platforms. For video, upload the highest-quality frame you can extract rather than a screen recording.
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.