How Scan works
Scan is a free, private document scanner. It runs entirely in your browser as an installable PWA — edge-detection, perspective correction, enhance, and OCR all happen on your device. Your documents never leave your device.
Your documents are never uploaded
There is no server and no backend. When you scan a photo, Scanruns OpenCV.js in your browser to detect the document edges, flatten the perspective, and enhance the image; optional OCR runs Tesseract locally; and the PDF is assembled by pdf-lib. The photo and the recognized text never leave your browser. The site's Content-Security-Policy (connect-src 'self' — the strictest in our suite of tools) even restricts which hosts the page may contact, so the privacy guarantee is enforced by the browser itself.
Vendored engines — no model CDN
Both heavy engines are vendored same-origin: OpenCV.js is served from /opencv/opencv.js and the Tesseract worker, WASM core, and English model live under /tess. They are fetched once, then cached for offline use. Because they are same-origin, the default CSP needs no external host at all.
Auto-detect with a manual fallback
Auto edge-detection usually finds the document, but it can miss on busy backgrounds or low contrast. When it does, you can drag the four corners to correct the selection manually — or keep the original photo unchanged. A bad auto-detect never blocks you.
OCR accuracy varies — be honest
OCR runs on your device with Tesseract. It is good, but accuracy depends on image quality, lighting, and font — a flat, well-lit, high-contrast scan (try the B&W document mode) recognizes far better than a skewed, dim photo. We make no guarantee that the recognized text is perfect; always review it before relying on it.
Camera access
On mobile you can capture a document directly with the camera. The Permissions-Policy allows the camera for this page only (camera=(self)) — a deliberate exception, because capture is core to a scanner (our other tools disable it). The photo is processed locally and never streamed anywhere.
No ads, no tracking, no account
There is no signup. This build ships with zero third-party scripts and a strict privacy policy. Monetization and Pro, if added later, will be environment-gated and off by default — the on-device privacy promise for your documents is never traded away.