How to scan a document with your phone

Why a photo is not a scan

A photograph of a page is taken at an angle, so the sheet comes out as a trapezoid rather than a rectangle. It carries the colour of whatever light you were standing in, the shadow of your own hand or head, and whatever happened to be on the table around it.

Turning that into a scan means four separate corrections: find where the sheet actually is, warp it back to a rectangle, scale it to a real paper size, and even out the lighting so the paper reads as white. ScanPDF does all four, which is why the output is closer to a flatbed scan than to a cropped photo.

Taking a photo that is easy to straighten

Automatic edge detection works by looking for a four-sided shape that contrasts with its surroundings. A few habits make that almost always succeed:

• Put the page on a contrasting surface. White paper on a dark desk is the easy case. White paper on a white table is the hard one, because there is no edge to find.
• Get all four corners in the frame, with a small margin around the sheet. A corner cut off by the frame edge cannot be detected, and cannot be dragged back later.
• Shoot from above rather than from the side. Any angle is correctable, but a steep one stretches the far end of the page and costs you sharpness there.
• Use even, indirect light. Turn the flash off. A window to the side beats an overhead lamp directly behind you, which puts your own shadow on the page.
• Flatten the page. A curl near the spine bends the edge, and a bent edge is harder to fit a straight line to.

None of this has to be perfect. Detection that misses is a five-second fix, and the guide below covers it.

Step by step

1. Add the photos. Drop them onto the page, click "Add photos", or use "Take photo" on a phone to shoot straight into the app. You can add a whole stack at once - each becomes one page.
2. Check the detected corners. Each thumbnail shows the frame that was found. Any page where detection was not confident is flagged with an "adjust corners" badge, so you only need to look at the ones that need you.
3. Drag the corners to the edges of the sheet. A magnifying loupe follows the handle so you can land precisely on the corner. "Re-detect" runs the automatic pass again, and "Reset corners" puts them back at the full frame if you want to start over.
4. Rotate if needed. Phone photos carry their orientation in EXIF metadata and are turned the right way up automatically, but the 90-degree buttons are there for pages photographed sideways.
5. Pick a filter, then use "Apply to all" if the whole batch was shot in the same light.
6. Order the pages. Drag a thumbnail to move a page, and delete the ones that did not work out. The order of the thumbnails is the order in the PDF.
7. Choose a page size and save. Auto, A4, Letter, a bank card, a photo print or another format, then "Save PDF". The file is assembled in the browser and lands in your downloads folder.

Choosing the right filter

The filter decides what the finished page looks like, and the right choice depends on what is on the paper.

• Color keeps everything as photographed. Use it for anything where colour carries meaning: highlighted text, stamps, signatures in blue ink, receipts, photographs on the page.
• Grayscale drops the colour cast from the room light without touching contrast. It is the safe middle option for mixed pages, and for anything containing a photograph you want to keep smooth.
• Black & white is the one that makes a document look properly scanned. It is not a hard threshold - it divides the image by an estimate of the background, which flattens shadows and uneven lighting, turns the paper genuinely white, and leaves the text anti-aliased instead of jagged. For plain printed or handwritten text it produces the cleanest, smallest file. Avoid it for photographs, which it will crush.

It costs nothing to try: switch filters and watch the live preview, since the choice is stored per page and only applied at export. A shadow across the page or the grey band along a book's spine has its own switch next to the filters: it evens out the lighting under any filter, and is best left off for pages that are mostly photographs.

Page size and output quality

A4 and Letter force every page to that exact size, which is what you want for anything that will be printed or filed alongside other documents. Auto sizes each page to the aspect ratio of the sheet you actually photographed - the right choice for receipts, tickets, ID cards and other odd shapes that would otherwise sit stranded in the middle of an A4 page. Auto is the default. The list also has exact formats for bank and ID cards, passports, business cards, photo prints, envelopes and more - pick one when the PDF page has to match the size of the real thing.

Detection runs on a downscaled copy for speed, but the export always warps the full-resolution original, capped at 3500 px on the long side and written at roughly 220 DPI on A4. That is comfortably enough for reading, printing, and for OCR later if you want it.

If the file has to stay under a limit - an email attachment, an upload form that rejects anything over 2 MB - pick a maximum file size next to the page size. ScanPDF first lowers the JPEG quality and then, only if it has to, the resolution, and it stops before text becomes unreadable. If the limit cannot be met that way, the smallest readable version is saved and you are told its actual size.

When detection cannot find the page

Automatic detection fails when there is no edge to find, and the cause is almost always one of three things: the sheet is on a surface of the same colour, a corner is outside the frame, or a hard shadow is cutting across the page and reading as a false edge.

The first two need a new photo. The third usually does not - drag the four corners in manually and the perspective correction works exactly as it would have otherwise. Detection is a convenience, not a requirement.

Nothing is uploaded

Every step above happens on your own machine. The images are decoded, analysed, warped and packed into a PDF by code running in your browser tab; there is no server to send them to. That matters for contracts, medical records, bank statements and ID documents, and it is explained in more detail on the privacy page.

Open the scanner and try it with whatever is on your desk.