Scan a document
Take photos, drop them here, and the four corners of each page are found for you — adjust any that are off, then turn the set into a clean PDF that can be made searchable. The photos never leave your device.
Drop document photos here or click to choose
JPG · PNG · WebP — corrected on your device
Drag a corner onto each corner of the document, or drag the handle in the middle of a side to move that whole edge. Pinch or use the zoom buttons to work closer in, and any handle can also be nudged with the arrow keys. The output is straightened into a rectangle.
Leave this off and nothing is downloaded for it. Turn it on and the OCR engine plus the English model — about 4 MB — arrive from shrink.run the first time, then read every page here and write an invisible text layer under the picture. English only: Turkish text is read poorly, and the picture you see never changes.
Auto flattens the light across the page, so the paper comes out white and the colours on it stay. Grayscale does the same and then takes the colour out. Black & white decides light or dark separately for every part of the page, which keeps small text readable. Original is the crop and the perspective correction and nothing else — the one to reach for when the thing you photographed is not paper.
Processed preview
This is the exact image written into the file you download.
How to scan a document
- Take clear photos or choose them from your gallery — one page or several.
- The corners are placed for you when the page edges are clear. Check them, and drag any corner — or any side — that is off.
- Reorder, rotate or remove pages until the set is right.
- Choose Auto, Grayscale, Black & white or Original; choose whether the PDF page is the size of the photo, an A4 or Letter sheet, or an ID card front and back on one page; and choose one PDF or a JPG for each page.
- Tick Make the PDF text-searchable if you want to be able to search the finished document.
- Create, check the processed preview and download.
The automatic corners are a starting point, never the last word. When the four sides of the page stand out clearly they are placed for you; when they do not, nothing is moved and the handles stay where they were. Either way you can correct them: dragging a corner opens a magnifier beside it, so the edge is not hidden under your own finger, the handle in the middle of each side slides that whole edge at once, and pinching or the zoom buttons let you work close in on a phone. Every handle also moves with the arrow keys, and Find edges again runs the search a second time.
Why a photograph is not a scan yet
A scanner puts its own even light on the page, which is the only reason scanned paper comes out white. A phone photographs whatever light is in the room: it falls off across the sheet, it takes the colour of the bulb, and the corner under your hand ends up much darker than the one by the window. That is why a straightened photo still reads as a photo. Auto estimates how bright the paper is in every part of the picture — ignoring the ink, and stepping over things like a passport portrait that are dark but are not shadow — and divides that back out, so the paper lands at white everywhere and the print keeps its colour. Black & white then asks the light-or-dark question separately in each neighbourhood rather than at one level for the whole page, which is what keeps a photograph inside the document from collapsing into a black blob.
It is worth saying what that assumes: that there is paper. Photograph a dark cover, a screen or an object and Auto will still try to make it white. Original applies none of this and is the right choice there.
Searchable scans, still without an upload
A photograph of a page is pixels: a PDF reader cannot search it, and neither can you. The searchable option runs the English OCR model here, in this tab, and writes what it read as an invisible text layer sitting exactly over the words in the picture. Find, select and copy then work, and the page looks identical because the layer is drawn in the PDF mode that paints nothing.
It is worth being plain about the limits. The reading is English only, so Turkish and other languages come out poorly. A blurred or badly lit photo produces poor text, and words the model was unsure of are left out rather than guessed at — the tool tells you how many pages ended up with text. And the engine plus its model, about 4 MB, download from shrink.run on first use, which is why the option is off until you ask for it: leave it alone and not a byte of it is fetched. If you only want the text and not a PDF, Extract text from an image is the smaller tool for that.
A scanner without an upload
Receipts, contracts and identity documents are exactly the files that should not be sent to an unknown conversion server. shrink.run performs the perspective correction and the PDF creation in this tab. A multi-page contract can be photographed page by page and saved as one document without any of those photos leaving the device; if you also have PDFs to add around it, combine the results with Merge PDF.
Frequently asked questions
Are document photos uploaded?
No. Each photo is decoded, straightened and written into the finished file inside your browser.
Can I scan several pages?
Yes. Choose or drop as many photos as you need, in one go or a few at a time. Every page gets its own corners and its own rotation, the order can be changed with the Move buttons or by dragging a card, and the whole batch is saved as one PDF.
How does it find the corners?
When a photo is added, the page is looked for automatically: a small grey copy is scanned for the strongest four-sided outline in it, and the handles are placed there. It is a suggestion, not a decision. If the outline is not convincing — a pale page on a pale desk, or a page running off the frame — nothing is moved and the handles stay where they were, because a confident wrong crop is worse than no crop. Once you move a handle yourself, nothing automatic will move it back.
Can the PDF be searchable?
Yes, with the Make the PDF text-searchable option, which is off unless you turn it on. It reads every page with the English OCR model on your device and writes the words as an invisible layer under the picture, so Find and copy work in a PDF reader. The picture itself is unchanged. English only — Turkish text will be read poorly — and the engine and model, about 4 MB, are downloaded from shrink.run the first time and then cached. With the box unticked, none of that is fetched at all.
Can I get images instead of a PDF?
Yes. Set Output to a JPG for each page and every straightened page is offered as its own JPG download. Choose One PDF to get a single document instead. The searchable option belongs to the PDF, so it is hidden when you ask for JPGs.
What do the four looks do?
Auto is the one to start with. It works out how the light fell across the page and divides it back out, so the paper comes back white in the shaded corner as well as the lit one, and the colours printed on it stay where they were. Grayscale does the same and then takes the colour out. Black & white decides light or dark separately for every part of the page instead of at one fixed level, which is what keeps small text readable and stops a shadowed corner turning into a solid block. Original is the crop and the perspective correction and nothing else — the honest option when the thing you photographed is not paper, because everything else here assumes it is. The processed preview shows the exact image that goes into the download.
Can I put the front and back of an ID card on one page?
Yes. Photograph both sides, put them in that order, and set Page to ID card. Both sides are drawn at the same width on one A4 page, the front above the back. It is meant for exactly two pages: with one you get the front alone on the same sheet, and with three or more the tool says so rather than guessing which two you meant.
Does the PDF page have to be the size of the photo?
No. Fit to document, the default, makes each page exactly the size of the straightened picture — right for a full page, and byte-for-byte the file this tool produced before the choice existed. A4 and Letter centre the picture on a real sheet with a half-inch margin instead, which is what you want if the result will be printed or filed alongside other documents; the sheet turns to follow a landscape page.
Does this open my camera?
The site does not request camera permission. On a phone, the system file picker may offer Take Photo alongside the gallery.
How many photos can I add at once?
There is no fixed cap, but every page is straightened in this tab, so the practical limit is your device’s memory. Photos are held as small previews until you press Create scan, and only one page is rendered at full resolution at a time.