Face swapping is the technique underlying most deepfake intimate imagery. Tools like DeepFaceLab, FaceSwap, DeepSwap, and dozens of mobile apps allow anyone to composite a face from one image onto the body of another. The resulting content can be photorealistic enough to deceive casual viewers and even moderate-sophistication investigators. Detection requires examining the boundary where the swapped face meets the original content, as well as the consistency of lighting, skin tone, and texture across the composite.

Visual signs of face swapping

Face-swapped images often show characteristic artifacts at the face boundary: a slight softness or blur at the hairline and jaw edges, inconsistent skin tone between the face and neck, mismatched lighting direction between the face and body, and subtle color temperature differences between the grafted face and the surrounding image. Video face swaps additionally show temporal instability, the face boundary flickers or shifts between frames, blinking patterns are inconsistent with head movement, and the face may briefly show the original features when the head turns rapidly.

Metadata detection for face swaps

Face swaps applied to existing photographs inherit the metadata of the base image. This means a face swap on a genuine photograph may still contain valid camera EXIF, making metadata analysis less useful than for fully-generated AI images. The base image's metadata may include camera make, model, GPS, and timestamp, all authentic, while the facial content has been completely replaced. Sensor noise analysis is more useful: authentic camera regions will show normal noise distribution, while face-swapped regions often show different noise characteristics from the generation process.

Mobile face-swap apps

Consumer face-swap apps (ZAO, Reface, DeepSwap, FacePlay, and many others) generate output at specific resolutions and with characteristic compression artifacts. Some mobile app outputs include metadata referencing the app. Most apply post-processing that smooths the face boundary, making detection harder. The most reliable signal for mobile face-swap outputs is the compression pattern: face-swap apps typically apply aggressive JPEG compression that creates specific artifact patterns at the face boundary.

What to do if you find a face-swapped image of yourself

Face-swapped intimate imagery is covered by the TAKE IT DOWN Act regardless of detection confidence. You do not need to prove the image is synthetic to assert removal rights, you need to assert that it depicts you in an intimate context without your consent. Upload your face photo to ScanErase for biometric scanning across 200+ platforms, then authorize 48-hour removal notices to all identified platforms.

Frequently asked questions