Analyze viral photos for generative diffusion artifacts, compression layer mismatches, edge discontinuities, and synthetic skin blending—entirely in your browser.
Generative diffusion models like Midjourney, Flux, and Stable Diffusion leave mathematical fingerprints that differ fundamentally from physical camera sensors.
Real cameras compress an entire scene uniformly under standard JPEG block matrices. When synthetic faces or manipulated elements are inserted or generated, their frequency residuals diverge sharply from background surfaces upon recompression.
Diffusion models denoise images iteratively. This often creates unnatural transitions where subject contours (collars, hair strands, ear lobes, jewelry) transition directly from hyper-sharp contrast to liquid-smooth blur with no natural focal depth of field.
Physical silicon CMOS sensors imprint microscopic Photo-Response Non-Uniformity noise across every capture. AI generated images lack authentic high-frequency hardware noise, exhibiting uncanny digital cleanliness broken only by synthetic texture prompts.