Hey, I'm Oktay, building FakeRadar forensic AI image & video detection. We go live here on July 5.
While we wait: drop an image you're not sure about (or one that genuinely fooled you), and I'll run it through FakeRadar and post back what the forensic layer actually shows ELA, frequency artifacts, face-swap signals, C2PA. Curious which ones are hardest to call.
the signals-not-verdicts framing is genuinely refreshing, especially for journalists who need to show their work without overclaiming. love that you surface per-face analysis and C2PA checks side by side rather than hiding them behind a paywall.
Ran a few AI-generated images through it and liked how it broke down each detection engine's signal instead of just slapping on a yes/no label. The face-swap breakdown was especially useful, feels like something I'd actually bookmark for work.
Useful after getting burned by AI swaps in my group chats. The per-face analysis flagged a tampered region in a screenshot I'd been ready to share, which felt like a real catch, and the explanation readout made it clear what signals tripped it.
Love that you're showing the signals instead of dropping a flat verdict — that transparency about uncertainty is exactly what journalists and fact-checkers actually need when defending a finding.
genuinely love that you surface signals instead of throwing a verdict at people, that choice alone makes the whole thing feel more honest than the usual AI or not binary you see elsewhere.