Open the camera or pick photos, let the page find the four corners (or drag them yourself), straighten the page and lift the shadows — then save every page as one PDF or as JPEGs, entirely in your browser.
Your files never leave your browser — all processing is local.
Drop photos of your pages here
or tap to choose photos — several at once. JPEG, PNG, WebP and HEIC. Nothing is uploaded.
Paper size (all pages)
Auto uses A4 or Letter when a page's shape is within 3 % of it, otherwise the page keeps its own shape at 300 dpi.
No pages yet. Open the camera or choose photos to start.
A photo of a contract, a receipt or a page of lecture notes is rarely usable as it is: the page is a trapezoid, the desk shows around the edges, and your hand or the phone itself casts a grey shadow across the text. This page turns that photo into a scan. Open the camera on your phone (the back camera, with the document outline detected on the live view) or pick photos you already took — many at once. For each one the page finds the four corners, straightens the perspective into a flat rectangle, lifts the shadow so the paper comes out white and the ink black, and adds it as a page. The pages go out as one PDF or as individual JPEGs. Nothing is uploaded: detection, warping and filtering all run in a worker inside your browser, which is also why this online document scanner works without installing a scanner app.
Corner detection is a classic edge-and-contour method, not a model, so there is nothing to download — and it has a known weak spot: when the paper and the desk are close in colour, white paper on a white or light desk, it often fails or grabs the wrong shape. When that happens, drag the four handles yourself; a magnifier above your finger shows exactly where the corner sits. A dark desk gives the best results. The correction is a flat perspective transform: it is right for a sheet lying flat and only approximately right for a book opened across the spine or a curled receipt — the page is straightened, but the curve is not flattened out. Paper size is one setting for the whole scan: auto reads the ratio of each page and snaps it to A4 or Letter when it is close, A4 and Letter force that size, and free keeps the detected ratio and sets the PDF page to the image at 300 dpi. If pages in auto land on different sizes, the whole PDF falls back to free so page sizes do not jump.
Each page has its own filter. Original keeps the colours after the perspective fix. Enhance, the default, estimates the uneven lighting across the page and divides it out, so a shadow from your hand or a lamp on one side becomes an even white background, then stretches the contrast; colours stay, so a red stamp or a highlighted line survives. Grayscale drops colour but keeps the tones. Black and white turns every pixel into either ink or paper, which gives the smallest, sharpest result for plain printed text — and loses everything that is not high-contrast ink: photographs, light pencil, a pale blue signature and most stamps simply disappear. For a document that carries a stamp or a signature you need to keep, use enhance, not black and white.
A scan can hold up to 50 pages. Each page can be moved up or down, rotated in 90-degree steps, deleted, or retaken from the camera or replaced by another photo, and each one keeps its own corners and filter. Very large photos are scaled so the long edge is 3200 pixels before processing, which is more than enough for an A4 page at 300 dpi and keeps a 12-megapixel phone photo from eating the browser's memory. Output is one PDF named scan.pdf, built by the same code as /images-to-pdf with your paper setting, or one JPEG per page named scan-01.jpg, scan-02.jpg and so on, with the rotation baked into the pixels. The PDF contains images only — this page does not recognise text, so the result cannot be searched or copied. To pull the text out of a page, send it to /image-to-text; to add a signature to the finished PDF, open it in /pdf-sign.
Yes. Open this page in your phone's browser, tap the camera button and allow camera access; the back camera opens with the document outline detected on the live view, and a tap captures the page. The camera needs a secure (HTTPS) page and your permission, and works in Safari on iOS as well as Chrome, Firefox and Edge on Android — it is not limited to Chrome. If the browser has no camera, or you deny access, the page says why and you can still pick photos from your gallery.
Drag them. Automatic detection looks for the edge between paper and background, so it often fails when the two are close in colour — white paper on a white or light desk is the usual case — or picks up a notebook, a keyboard or a table edge instead. When nothing is found, the frame starts slightly inside the edges of the photo. Pull each handle to its corner; the magnifier above your finger shows the exact spot. A dark, plain desk under the paper is the single thing that helps most.
That is what the enhance filter is for. It estimates how the lighting changes across the page — a hand shadow on one side, a lamp on the other, a yellow cast from indoor light — and divides it out, so the background becomes an even white and the text stays dark, with the colours kept. It works on soft, gradual shadows. A hard, sharp-edged shadow, or one so deep that the text underneath is nearly invisible in the photo, will leave a visible mark; retaking the photo with the light in front of you beats any filter.
No. The correction is a flat perspective transform: it assumes the page is a flat rectangle seen at an angle. A book open across the spine or a receipt that curls will be straightened at the corners, but the lines of text will still bend where the paper bends. Pressing the page flat before you shoot, or scanning one half at a time, gives a better result than any setting here.
Because the page is on the black-and-white filter. It decides, pixel by pixel, whether something is ink or paper, which is ideal for crisp printed text and bad for everything in between: light pencil, a pale signature, photographs and most stamps fall on the paper side and vanish. Switch that page to enhance, which removes the shadow but keeps colours and tones. The filter is per page, so a form with a stamp can use enhance while the plain pages stay black and white.
No. This page does not recognise text, so the PDF it writes contains images only, with no text layer — it looks like a scan and behaves like one. To get the words out of a page, send the same photo or the scanned JPEG to /image-to-text, which runs OCR in the browser and gives you the text to copy.
No. The camera stream, the photos you pick, the corner detection, the perspective warp, the filters and the PDF itself all stay in your browser; the work runs in a web worker and the download is built from memory. There is no server behind this page, which matters precisely because the documents you scan are contracts, receipts and homework with names on them.
One setting for the whole scan. Auto reads each page's ratio after correction and uses A4 or Letter when it is within a few percent of one; A4 and Letter force that size with the page scaled to fit and no margin; free keeps the detected ratio and makes the PDF page the size of the image at 300 dpi, which is the right choice for receipts and odd sizes. If auto lands on different sizes for different pages, the whole PDF uses free so page sizes stay consistent. The filter, by contrast, is chosen per page.
Pull the text out of a scanned page with OCR, in your browser.
Already have clean images? Combine them into one PDF with your own page size and margins.
Add your signature to the finished scan.pdf before you send it.