Chrome extension + local Python server for on-device text / image / video authenticity checks. Public repo: alexnstevens06/LucidScan. Built for a Fall 2025 hackathon (repo created 2025-10-25 CT; single substantive commit: “all files at end of hackathon”).
What it is
chrome_plugin/— extension that sends selected text or media URLs to a local/detectendpoint and shows a confidence score in a popup.server.py— Flask API that routestext/image/videopayloads to detection helpers.text_detection.py/detect.py— text classifier path plus CLIP/ViT-based image and video heuristics (watermarks, EXIF, Gradio GUI option).
The checked-in root README is still largely an upstream StableGuard template (clone instructions point at muhammad-fiaz/StableGuard). Treat that marketing block as scaffolding, not the project description. This write-up describes what is actually in the LucidScan tree.
Honesty / scope
Thin history: one push at hackathon end, not a maintained product. Suitable as a systems/ML side project showing extension ↔ local-model wiring, not as a flagship. Resume claims (fine-tuned models, sub-5s local runs) are aspirational relative to the public commit surface — the code loads open models and runs inference stubs; there is no separate fine-tune artifact in-repo.
Tech
Python (Flask, PyTorch / Transformers / Gradio), JavaScript Chrome extension (popup + background).
Dates
October 2025 (hackathon).
Links
https://github.com/alexnstevens06/LucidScan (public)
Figures
None. The repo has no usable screenshots or UI captures without third-party / StableGuard branding. Cover left empty for Castellan.