Clean up scans and document photos

Compression makes the file smaller and leaves the page looking exactly as it did: grey stays grey and a tilted page stays tilted. This tool works on the image itself, whitening the paper, removing the lighting gradient and the shadow, and correcting the tilt. The file usually gets much smaller too, because flat white compresses far better than a grey gradient, but that is a side effect rather than the point.

No. The result is clean page images, not selectable text. If you need text you can copy or a searchable file, run the result through the OCR tool afterwards, where it will read more accurately on a straight, high-contrast page than on the original photo.

Up to six degrees either way, which is the range a hand-held photo lands in. A page that is upside down or a quarter turn out is a rotated page rather than a crooked scan, and the rotate tool is the right one for that. The angle is measured from how the text lines align, so a page with no text is left exactly as it is.

White paper is the default: a white background with grey-shaded ink, and it is the right one for Arabic because the shading keeps the edges of letters and their diacritics. Keep colour whitens the background but leaves blue and red stamps and signatures in colour. Black and white forces every dot to one or the other, which gives the smallest file and the sharpest contrast but can swallow very fine strokes.

Detail does not increase; nothing brings back what the camera did not capture. What improves is legibility: the separation between ink and paper becomes sharp instead of a grey ramp. Pages are processed at up to 2400 pixels on the long edge, which is enough for printed text and keeps the work possible on a phone.

No. All of the processing happens inside your browser on your own device, and the images and files never leave it at any step. You can disconnect from the internet after the page loads and the tool still works.