What it does Generates clean base QR codes, then compares them pixel-by-pixel against AI-generated variants to flag exactly where the diffusion model mangled the pattern. There’s a walkthrough guide and a live web app, plus a companion extension for AUTOMATIC1111’s Stable Diffusion Web UI.

The interesting bit The hard part of AI QR codes isn’t making them pretty—it’s making them scannable. This toolkit treats QR readability as a diffing problem, which is a pleasantly mechanical solution to an otherwise fuzzy generative-AI headache.

Key highlights

• Live app at qrcode.antfu.me — no install required
• Pixel-level comparison between base and generated QR codes
• Companion SD Web UI extension for in-pipeline refinement
• Vue-based frontend; MIT licensed
• Detailed walkthrough by Anthony Fu on his personal blog

Caveats

• README is sparse on technical architecture; you’ll need to read the code or blog post for implementation details
• The SD Web UI extension lives in a separate repo, so this core repo is just the comparison toolkit and web app

Verdict Grab this if you’re iterating on AI-generated QR codes and tired of waving your phone around hoping one scans. Skip it if you just need a standard QR generator—there are lighter tools for that.