Back to browse
GitHub Repository

Local-first inspection engine to detect AI-generated video artifacts, deepfakes, and diffusion slop.

0 starsJavaScript

Detecting deepfakes without sending your files to a cloud API

by XQorp·May 11, 2026·2 points·0 comments

AI Analysis

●●●BangerBig BrainWizardry

Detects deepfakes locally using optical flow vectors instead of sending files to cloud APIs.

Strengths
  • Uses Farneback's algorithm to map pixel vectors rather than relying on black-box neural nets.
  • Fully offline architecture ensures sensitive media never leaves the user's machine.
  • Visualizes temporal incoherence via 36-bucket directional histograms for explainable results.
Weaknesses
  • Optical flow alone may struggle with high-quality diffusion models that maintain object permanence.
  • Requires local Python and OpenCV setup, creating friction compared to drag-and-drop web tools.
Category
Target Audience

Security researchers, journalists, and privacy-conscious users

Similar To

Hive Moderation · Sensity AI · Microsoft Video Authenticator

Similar Projects