Clean Document Scan

This tool takes a single document photo or scan and makes the text easier to read by adjusting brightness and contrast, dropping the color cast through grayscale or black-and-white conversion, and sharpening soft edges. It fits the everyday gap between a phone snapshot of a page and something you would actually file, print, or run through OCR.

Realistic uses cluster in a few places. Office and admin: brightening a dim scan of a letter, form, or receipt before it goes into a records folder. Students: cleaning up a photo of a handout or library page so it is legible at a glance. Sharing: lifting a washed-out or yellow-tinted scan to a crisp B&W copy that reads well in email or chat. Data prep: as a pre-pass before an OCR tool, since higher contrast and reduced color noise generally help text recognition.

It is deliberately narrow. There is no document classification, no text extraction, and no multi-page handling here. One image in, one cleaned image out, with a before/after slider so you can judge the change rather than trust a label.

Drop or select an image, choose a mode, and click Clean scan. Auto mode applies a balanced preset tuned for printed pages: a modest brightness lift, a stronger contrast boost, moderate sharpening, and grayscale conversion. Manual mode exposes the same controls directly: brightness and contrast each from -80 to +80, sharpness from 0 to 100, and a color mode choice of Color, Grayscale, or Black & White. All processing happens in your browser on an HTML canvas; on this site the file is not uploaded and nothing persists once the tab closes.

Inputs are JPG, PNG, and WebP up to 30 MB, one file per run. The export is a PNG named after your original with a -cleaned suffix; despite what some listings imply, this tool writes PNG only, so there is no direct JPG or PDF save here. Sharpening is an unsharp mask (a blurred copy subtracted from the original), and Black & White is a single global luminance threshold at the midpoint rather than an adaptive, region-aware one.

What it does not do matters as much. It will not deskew a tilted page, correct perspective on an angled photo, auto-crop to the document edges, or even out shadows from uneven lighting. Those are handled by separate scan-cleanup tools (Deskew, Auto Crop, Remove Shadows). Removing yellowing here means the color cast disappears when you convert to grayscale or B&W, not a true white-balance correction that keeps color.

The most common disappointment is pushing contrast or the auto preset too hard: thin or faint text can break up, and light pencil or low-ink print can vanish into the background. Over-sharpening is the mirror problem; past a moderate setting it amplifies paper grain and JPEG artifacts into visible speckle. When that happens, switch to Manual and ease the sliders back rather than re-running Auto.

Black & White is the sharpest trap. Because the threshold is fixed at the midpoint and applied globally, a scan with uneven lighting or a shadowed corner will turn that whole region solid black or blank white. If your photo has shadows or a light gradient across the page, clean the lighting first or stay in Grayscale, where detail is preserved. B&W is also wrong for anything with photographic tone, gradients, or colored highlighting you need to keep.

A few smaller pitfalls: the tool cannot fix motion blur or focus blur, only soften-to-sharp edge contrast, so a genuinely out-of-focus shot stays unreadable. It also will not straighten or crop, so a crooked, cluttered photo comes out cleaned but still crooked and cluttered. And output is always PNG, which for a large color page can be a heavier file than the JPG you started with.

Use the before/after slider and read the worst part of the page, not the best. Confirm the smallest text is still fully formed, that no letters have been eaten by contrast, and that the background is clean without new speckle from sharpening. If you chose Black & White, scan the corners and any shadowed areas specifically for blown-out or filled-in regions, since that is where the global threshold fails first.

If the scan is headed for OCR, judge legibility rather than prettiness: even, well-separated characters matter more than maximum darkness. If it is headed for print or sharing, check the PNG file size is reasonable and that grayscale or B&W did not strip something you needed, such as a colored stamp, signature ink, or highlight.

For anything official or financial, treat the cleaned image as a more readable copy, not a verified one. Always compare it against the original document, confirm no figure, signature, or character was distorted by the processing, and keep the untouched source as the record of truth.