How to scan documents with your phone, and what 'scanning' actually does
Phone cameras killed the home scanner, and "scanning" now means a photo of a page put through software. The software is what makes the difference between a crooked, grey snapshot and a document a bank will accept — and it is simple enough to run in a browser, as the document scanner does. Here's what it's doing, and how to shoot pages so it does it well.
Why a photo isn't a scan
A flatbed scanner sees the page flat, evenly lit, at a known size. A phone sees it from an angle (so the rectangle becomes a trapezoid), under uneven light (a brightness gradient across the page, shadows from your hand), in colour (the paper is grey-yellow, the "white" varies), and at whatever size the framing happened to give. A scanning app corrects each of those; a raw photo corrects none, which is why forms photographed and emailed get bounced back.
What the app does: four steps
- Find the page. Edge detection looks for the four corners; when it guesses wrong (patterned tablecloth, low contrast), you drag them. Good apps let you; browser tools ask you to, which takes ten seconds and is always right.
- Straighten. The four corners define a perspective transform — mathematically a homography, a 3×3 matrix mapping the trapezoid back to a rectangle — and every pixel is resampled through it. The output size comes from the average of opposite sides, so proportions are preserved.
- Clean. Auto levels stretch the paper to white and the ink to black; a "document" or black-and-white mode thresholds the page so text becomes crisp and file size drops sharply. Colour mode keeps stamps and signatures.
- Package. Pages are fitted to A4 or Letter and stacked into a PDF; the good apps then run OCR so the result is searchable.
Shooting the page so it works
- Put the page on a contrasting, plain surface — a dark table for white paper — so the edges are obvious.
- Light it evenly from the side or above; avoid your own shadow and the flash, which makes a hot spot.
- Hold the phone parallel to the page and fill the frame. Perspective correction can fix an angle; it can't add pixels to the far edge.
- Flatten the page. Curl and creases can't be corrected; a book spine will always shadow.
- For text-only pages choose black & white: smaller and sharper. For anything with a photo, colour stamp or signature, keep colour.
Output: PDF, size, searchability
A page at 150–200 dpi in black and white is 50–150 KB; in colour, 300–600 KB — a 20-page contract fits in an email either way (What DPI you actually need for printing explains the resolution arithmetic). The pages are pictures, so the PDF isn't searchable until OCR adds a text layer — the searchable PDF tool does that on-device, and How OCR works, and why scans read better than photos explains what it reads well. For bulk archives, scan black-and-white and OCR everything; for anything you'll sign or submit, keep a colour copy too. The PDF compressor shrinks the results further.
Where your documents go
The pages people scan are the sensitive ones — IDs, contracts, bank letters, medical forms — and the popular scanning apps upload them to cloud OCR services, sometimes with a subscription attached and always with terms of service. The phone's built-in scanner (Notes on iPhone, Google Drive's scan on Android) stays local. So does a browser tool that does the straightening, cleaning and PDF assembly in the page: the photo never leaves the device, there's nothing to install, and the output is a standard PDF. For a contract, that's the version to use.
Sources and further reading
The claims in this guide rest on these references, which were checked when the guide was last updated. Spotted an error? The contact page says how to report it.