AI Deepfake Detector

Lightweight synthetic-media detector, no GPU

A lightweight deepfake and synthetic-media detector that runs on any device with Python and needs no GPU. It combines several independent checks: lighting consistency (conflicting light directions across regions suggest splicing or relighting), body and face anatomy from landmark geometry (eye, mouth, shoulder and hip symmetry, iris placement), mirror and reflection symmetry, EXIF metadata and software traces left by editors and image generators, and a re-compression heuristic for double-saved images. It works on photos and videos and can write a JSON report.

Languages
Built with
OpenCVMediaPipe
GitHub

Started Last updated

November 2026