Scan a document to PDF

It straightens, but it does not detect. Press Crop and four handles appear at the corners of the photograph; drag each one onto a corner of the page and the area inside is warped back into a rectangle, which is exactly what removes the trapezium shape you get from photographing a page at an angle. What is missing is the automatic part, deliberately. Edge detection is not hard to write badly — it is hard to write safely, and a page on a patterned tablecloth or with its bottom edge just outside the frame produces a confident, wrong result that looks like a clean scan and is missing a line. When you place the corners, you have seen where they are.

Original leaves the photograph alone. Document measures how bright the page is overall and stretches that range so paper comes out white and ink dark, which is what makes a phone photo read as a scan. B&W does the same and then forces every pixel to pure black or white, which suits printed text and makes the smallest file. Board keeps the colours so whiteboard pens survive while the background is lifted to white. Photo keeps the colour and lifts it gently, for a page where the colour is the point. All five are brightness, contrast and saturation arithmetic — none of them sharpens, removes shadows or invents detail, and one that would have to amplify a dark photo too far falls back to plain greyscale rather than turning noise into speckle.

Yes — choose Searchable PDF before building, or press “Make it searchable” on the result. Recognition reads the words and places them behind the pages invisibly, so the document can be searched and its text selected while looking exactly as it did. It runs on your device and needs a language model downloaded once, which is why it is offered rather than always on. The plain image-only PDF is built first either way, so a long recognition that fails cannot cost you the pages you photographed.

It is not. The video runs into a preview element in this tab and each captured frame is drawn onto a canvas, cleaned up and encoded, all on your device. Nothing carrying an image goes out — open your browser's network panel and scan a page if you want to confirm it. The camera is also released the moment you press Stop or leave the page, so the indicator light goes out when you are done.

There are four common reasons and the tool names which one it hit. The permission was refused, in which case allow it from the padlock or camera icon in the address bar. There is no camera on the device. Another application or tab is holding the camera, so close it. Or the page is not on a secure connection, which browsers require before granting camera access at all. In every case you can photograph the pages with your normal camera app and add the images instead.

Yes. There is a file picker below the camera that takes any images your browser can open, and they go through exactly the same processing and end up in the same PDF. That is also the way in for anyone whose camera cannot be used here — a desktop with no webcam, a blocked permission, or a work network serving pages without encryption.

Fifty. Each captured page is held at full sensor resolution until you build the PDF, and that is memory on the phone this is most likely running on. Past that the limit is enforced rather than let the tab run out of memory and lose the pages already taken. Build the PDF and start another batch, then merge the results.

The rear one, which is asked for as a preference rather than a requirement. Requiring it would fail outright on a laptop that only has a front camera and would otherwise work fine, so if no rear camera exists the available one is used instead.