Scan to PDF

How to scan a document without a scanner

Almost nobody owns a scanner any more, and almost everybody is carrying a better one. A phone camera photographs a page at more detail than a cheap flatbed ever managed. What the scanner apps add is the tidying up and the putting together, and both of those can happen here.

1. Open this page on your phone, or on a laptop with a camera.
2. Press Turn the camera on and allow it when your browser asks. On a phone the camera fills the screen.
3. Lay the page on a desk or a floor that is darker than the paper, in as even a light as you can.
4. Wait for the green outline to sit on the edges of the page. It says Page found when it has it.
5. Press the round button to take the page. Do the same for every page, in order.
6. On a phone, press Done when you have them all.
7. Remove any page that came out badly and take it again.
8. Press Save as a PDF.

What makes it a scan and not a photograph

A photograph of a page is taken from wherever you happen to be standing. The desk shows around the edges, the page is a little crooked, and because the camera is never exactly square to the paper the far edge comes out shorter than the near one. A scanner has none of those problems, because the page is pressed flat against the glass and nothing but the page is in the picture.

So the picture is measured before it is kept. The bright area in the middle is followed out to its edges, the four corners of the page are picked out of it, and the outline you see in the viewfinder is those four corners drawn back over the video. When you take the shot, everything outside them is thrown away and the four corners are pulled out to the four corners of a rectangle, which straightens the page and undoes the lean of the camera at the same time. What is left is the page, square, filling the whole picture.

It needs to be able to tell paper from what is under it, so a white page on a white table is the one thing it cannot find. If the outline will not settle, move the page onto something darker, or turn the finding off and take a straight photograph instead - the cleaning still happens either way.

What cleaning the page does

A photograph of a white page is not white. It comes out grey by a window, yellow under a bulb, and darker in one corner than the other. That is a problem of light, not of colour, so cleaning fixes the light and leaves the colour alone. The brightest tenth of the picture is taken to be the paper, everything is scaled so that the paper lands on white, and a little contrast is put back so the writing does not fade with it. Scaling the red, green and blue separately is what takes the yellow out of a page shot under a lamp.

A red stamp stays red, a highlighted line stays yellow, a logo keeps its colours, and the page behind them goes white. That is what people mean when they say a scan looks clean, and it is what you get here without asking for anything.

There are two other choices in the bar. Black and white drops the colour entirely and pushes the page to ink and paper - the smallest file, and the sharpest result for a long document that is nothing but text. Straight photo changes nothing at all, for a drawing, a photograph, or anything you need exactly as the camera saw it.

Why this does not want your photographs

The free scanner apps are free because of what they do with the pages. A document you scan is uploaded, held, and often used to sell you a subscription to get it back without a watermark. A payslip, a passport, a medical letter and a signed contract are the four things people scan most, and they are the four things you would least like sitting on a stranger's server.

Here the camera is opened by your own browser, the picture is drawn onto this page, and the PDF is assembled in front of you. There is no server to upload to. You can load this page, turn your internet off, and scan a whole pile of paper anyway.

The text is not read

A scan is a picture of a page. The words in it cannot be selected, searched or copied, here or in any other scanner, until something reads them back out - which is a separate job called OCR. This tool does not do that yet. If you need searchable text, say so and it can be built.