Verify Genuine Portraits & Spot Deepfakes
Synthesized faces and face-swapped portraits frequently exhibit subtle blending discrepancies around hair margins, iris asymmetry, and uniform dermal smoothing. Drop any portrait below to inspect optical and geometric authenticity.
- Target Forensics
- StyleGAN, InsightFace, FaceFusion, Flux Portraits
- Key Vectors
- Pupil symmetry, hairline boundary blending, ear symmetry
- Processing
- 100% Client-Side Canvas & Frequency Analysis
- Privacy Guarantee
- Zero facial data leaves your local device
Drop an image here
or choose a file · paste with ⌘V / Ctrl V
How Deepfakes and Face Generators Leave Unnatural Artifacts
Generative face synthesis architectures—including StyleGAN, FaceFusion, and recent diffusion-based face replacers—generate convincing human likenesses by mapping facial landmark geometries into high-dimensional latent vectors.
However, physical human biology obeys strict biological constraints that models struggle to maintain simultaneously: corneal light reflection angles, ear cartilage symmetry, hair strand continuity against backgrounds, and subsurface light scattering across cheek capillaries. Our detector inspects these telltale optical anomalies.
Corneal Catchlight Consistency
In genuine photography, light reflecting from both left and right corneas mirrors identical physical environmental light sources. AI-synthesized portraits frequently exhibit conflicting reflection angles and mismatched specular glints.
Hair-to-Background Blending Boundaries
Fine flyaway hair strands against natural or complex backgrounds are notoriously difficult for inpainting and swapping models, resulting in localized blurred halos or geometric continuity breaks.
Ear Cartilage & Jewelry Asymmetry
AI image generators treat left and right ears as independent probability fields, often producing asymmetrical tragus/helix shapes or mismatched, dissolving earrings.
Uniform Dermal Micro-Smoothing
Natural camera sensors capture organic ISO noise across facial skin pores. Generative models substitute natural noise with algorithmic pixel smoothing that fails Laplacian gradient variance tests.
Questions about Deepfake & Face AI.
Looking for technical verification guidance or have a specific question about Deepfake & Face AI artifacts?
Can this detector tell if a face is generated vs a real camera photo?
Yes. Our client-side forensics examine sensor noise, frequency spectrum distributions, and compression artifacts to deliver an instant AI likelihood reading without sending your photo to any server.
Are my personal photos or selfies uploaded to any server?
No. Check AI Free executes 100% in your browser memory via HTML5 Canvas. Zero pixels, filenames, or biometric data are transmitted across the network.
Does it detect face swaps on existing real photos?
Yes. Face swap algorithms leave distinctive boundary blending gradients and frequency mismatches between the replaced face oval and the original neck/hair tone.
Also inspect other generative models
Midjourney
Detect Midjourney v5 and v6 cinematic images
Flux.1
Detect Black Forest Labs Flux.1 Schnell, Dev, and Pro
DALL·E 3
Verify images generated by OpenAI DALL·E 3 and ChatGPT
Stable Diffusion
Inspect SDXL, SD 1.5, and ComfyUI generated artwork
AI Profile Picture
Identify AI-generated LinkedIn avatars, headshots, and social bios
Instagram AI Photo
Detect AI-generated virtual influencers, models, and filtered photos