See the reader actually read.

This page runs Tesseract.js OCR locally — your image never leaves your device. Same OCR family powers the open-source PlateHunt edge agent. Pick a sample plate or upload your own.

1 · Sample plates

Generated plates with known text — proves the pipeline works end-to-end.

2 · Or upload your own

📷 Drop a plate photo or click to choose

Live sighting feed (from the cloud API)

Every plate read above gets posted to /api/sighting and stored in Cloudflare D1. This feed pulls the last 50 from that table.

What this demo proves and what it doesn't:
✓ The OCR pipeline works in real time on real images.
✓ The cloud ingest API + storage works (D1 logs every read).
✓ The whole stack runs free-tier — your costs at this stage = $0.

Browser-Tesseract accuracy is ~75–85% on clean plates. Production PlateHunt swaps in a commercial-grade ALPR (PlateRecognizer or self-hosted CodeProject.AI) for 96%+. The pipeline architecture is identical — just the OCR backend differs.