The problem
Pulling text out of a pile of document photos by hand is slow, and most OCR tools keep what you upload.
What it does
Drop up to 100 images in and Aynscan extracts the text from each one, three at a time. Results appear as searchable cards: filter them live with highlighted matches, open one to copy its text, or clear the batch when you are done.
How it's built
- OCR: Llama 3.2 11B Vision on Cloudflare Workers AI, reached through a thin Cloudflare Pages Function that keeps the API token off the client.
- Client: React 19 + TypeScript + Vite. Images are downscaled in the browser before upload.
- Storage: none. Files and results exist only in browser memory, and the app installs as a PWA.
Notes on the demo
The video on this page uses mocked OCR responses so it can be recorded without credentials. The real app calls Workers AI, which means images do leave the device for Cloudflare's API; nothing is stored afterwards.