- category
- detector
- maturity
- emerging
- approach
- Browser extension that runs local inference (an ensemble of two Hugging Face vision classifiers) directly on images and video frames as a webpage renders, drawing red border overlays and confidence badges on content flagged as AI-generated, with no upload to any server.
- access
- open-source
- owns
- A privacy-first, in-page detector for AI-generated images and video frames, requiring a small local Python backend for offline inference; adjustable confidence thresholds, no accounts or telemetry.
- distinctFrom
- Unlike ai-or-not or hive-moderation (web-upload or API services users visit to check one file at a time), this inspects every image already rendered on whatever page a user is browsing, in real time, entirely on-device.
- notes
- MIT license, 7 stars at time of check; very small/early project, requires the user to run a local Python backend, so real-world adoption is minimal. Vendor-cited 85-95% accuracy range is unverified. Exact project age/launch date could not be confirmed; not previously tracked in this catalogue, flagged as an untracked find rather than confirmed as a brand-new 2026-07-27-window launch.
