Kit-Bin
Donate

Scan a Document to PDF

Processed entirely in your browser. Never uploaded to a server.

Drop a photo of a document here or click to browse

Processed entirely in your browser. Never uploaded to a server.

Drop a photo of a paper document above and this tool tries to find its four edges, straighten it into a flat, top-down rectangle the way a real scanner would, boost the contrast so it reads like a scan rather than a photo, and hand you back a single-page PDF. It's the free, offline version of the "scan to PDF" feature built into most phone camera apps.

What actually happens to your photo

The photo is decoded onto a canvas, edge-detected, and searched for the largest four-sided outline in the frame — the document's edges, on a well-lit shot against a plain background. When that outline is found with reasonable confidence, a perspective transform warps it flat, correcting the skew from photographing the page at an angle. A contrast boost runs afterward either way, so the result looks crisp and scan-like rather than like a photo of paper.

Getting a good result depends on the photo

This is the honest part: auto-detection is only as good as what it's given. Good, even lighting with no harsh shadow across the page, the document filling most of the frame rather than sitting small in a wide shot, and a background that contrasts with the paper (a dark desk under a white page, for example) all meaningfully improve corner detection. A cluttered background, a page that blends into the surface under it, or heavy glare can all cause auto-crop to miss — in which case this tool falls back to your original photo rather than producing a bad crop or failing outright.

FAQ

How does the auto-crop actually work?
It runs a classic computer-vision pipeline entirely in your browser: find edges in the photo, look for the largest four-sided shape those edges form, and if one is found with reasonable confidence, warp it flat so the page fills the frame corner to corner, like it was photographed head-on. This runs on-device, using an OpenCV build compiled to WebAssembly, not a server.
What if it picks the wrong edges, or doesn't find the page at all?
Then it falls back to your original photo, uncropped and unwarped, rather than failing outright or guessing badly. Being honest about that: a photo where the document doesn't fill most of the frame, or where the background is a similar colour to the paper, is the case most likely to trip up auto-detection. Good lighting and a plain, contrasting background behind the page both help a lot.
Can I adjust the corners myself if auto-detect gets them wrong?
Not in this version. There's no manual drag-the-corners step — it's auto-detect or the full photo, one or the other. If the auto-crop consistently misses on a particular photo, retake it with the document flatter, better lit, and filling more of the frame, or use Crop Image afterward on the resulting page.
Does this work on a whiteboard, a receipt, or a page from a book?
A whiteboard or a receipt against a contrasting surface usually works fine, since both are still a rectangular thing with clear edges. A book page is harder: the curve near the spine and the second, adjacent page in frame both confuse the four-corner assumption this tool makes, so expect the fallback (full photo) path more often there. It still produces a usable PDF, just without the perspective correction.
Can I scan a multi-page document in one go?
Not yet — this is a single photo in, single-page PDF out tool for now. Scan each page separately and combine the resulting PDFs with Merge PDF, or use Images to PDF if you'd rather skip the scan-look processing for some pages.
Is my photo uploaded anywhere?
No. The edge detection, perspective warp, and the PDF itself are all built in this browser tab. The only thing ever downloaded is the scanning library itself, a few megabytes, cached by your browser after the first use — your photo never leaves your device.

Related tools: Images to PDF, Compress PDF, Crop Image.