Straighten and clean up a scan

How it works

1. Drop the scans or photos of documents. Several can be cleaned in one run, each measured on its own.
2. Leave Straighten the page on to have the tilt found and corrected. The tool reads the rows of text: it tries every angle up to 10 degrees either way and keeps the one where the lines come out sharpest, to a tenth of a degree.
3. Leave Crop to the paper on if the photo shows the desk around the page. A scan that is already trimmed is left at its full size.
4. Choose the finish. Greyscale is the usual one for text. Black and white decides each pixel against the brightness around it, which keeps a page lit from one side readable and makes much smaller files. Keep colours suits a form with coloured stamps or a highlighted passage.
5. Press Clean up scan and download. Files keep their name with -scan added.

Frequently asked questions

How does it know which way the page is tilted?

From the text itself. Lines of text pile ink into rows, and those rows only line up at one angle, so the tool shears the page through every angle in the range and keeps the one that makes the rows sharpest. It needs text to work: a photo of a drawing or a blank page has no rows to read, and is left exactly as it is rather than turned by a guess.

Will it fix a page photographed at an angle?

Only the rotation, not the perspective. If you shot the page from above but held the phone crooked, this straightens it. If you shot it from the side, so the page is a trapezoid rather than a rectangle, that needs corner-by-corner correction, which this tool does not do; retake the photo from directly above.

Why did it not crop my scan?

Because it found no darker border to crop away. Cropping looks for where the bright paper starts and stops, so it works on a photo of a page on a desk. A flatbed scan that is already all paper has nothing to remove, and a page photographed on a white table has no edge to find.

Does it straighten Japanese or Chinese text set vertically?

No. The tilt is read from rows of text, and vertically-set Japanese makes columns instead, so the page is left as it is rather than turned by a guess. Rotate it yourself with Rotate an image; the cropping and the contrast still work on it.

Which finish should I use before OCR?

Greyscale or black and white, both of which make text stand out from the paper. Black and white also shrinks the file a lot, which helps if the scan goes into a PDF afterwards. Read text from an image handles the reading itself, in the browser, with no upload.

Does it upload my documents?

No. Everything is measured and redrawn in your browser, which matters for contracts, invoices and ID documents. There is no model to download either: the whole thing is arithmetic on the picture.

Can it clean a PDF scan?

For the tilt, yes — Straighten a PDF does it on the PDF itself, so the scan keeps its resolution and any real text stays selectable. This tool only takes pictures, so for cropping and cleaning up a PDF scan you still turn it into images with PDF to JPG, clean them here and put them back with JPG to PDF.

What are the free limits?

The free plan handles 3 images per run, 20 MB each and 10 runs per day. Pro removes these limits.