Deskew PDF
Straighten a crooked scan by measuring its tilt, or set the angle yourself.
Straighten A Scan Without Re-Scanning It
A page fed into a scanner half a degree off sits crooked in every copy made from it afterwards. It looks careless, it makes text harder to read, and it wrecks OCR accuracy — character recognition assumes text lines run horizontally, and a slight tilt is enough to confuse it.
This tool measures the tilt from the scanned pixels and applies the correction as a transformation on the page. Nothing is re-rendered: text that was selectable stays selectable, and an image keeps every pixel it started with.
Key features
- Measures the angle for you — using the projection profile method, the same approach OCR software uses before reading a page.
- Correction as a transform — applied to the page, not by rasterising it, so nothing degrades.
- Hundredth-of-a-degree control — with nudge buttons, because the last fraction is always judged by eye.
- A ruled preview — the eye reads a slope against parallel lines far better than against a number.
- Rotates about the page centre — so the correction does not also slide content off an edge.
This tool runs entirely inside your browser using native Web APIs. Your files and text are never uploaded to a server, never logged and never shared with third parties.
How to use: Deskew PDF
- Choose the scanned PDF
The tool reports how many pages carry embedded images, which is what makes measurement possible. - Detect the angle
The largest scanned page is analysed and the slider is set to the measured skew. - Check and adjust
Compare the ruled sheet against the true horizontal line and nudge if it still looks off. - Download the straightened copy
Every page is rotated about its own centre by the same angle.
Technical specifications
| Processing location | Entirely in your browser — no server round trip |
|---|---|
| Data uploaded | None. Files and text never leave your device |
| Price | Free — no account, no trial, no usage cap |
| Category | |
| Works offline | Yes, once the page has loaded |
| Browser support | Chrome 90+, Edge 90+, Firefox 90+, Safari 15+ |
| Interface languages | English, 中文, हिन्दी, Español, العربية |
Frequently asked questions
How does automatic detection work?
The page image is reduced, thresholded into ink and background, and then rotated through a range of candidate angles. At each one the dark pixels are summed row by row. When the text lines are horizontal every pixel of a line falls into the same row, which makes those sums extremely uneven — so the angle with the most uneven profile is the angle that straightens the page.
Why will it not detect the angle on my document?
Detection needs pixels. A PDF exported from a word processor draws its text as vectors and contains no image to measure, so there is nothing to analyse — and such a document is almost never crooked in the first place. Use the slider if you need a deliberate tilt.
Will straightening blur the text?
No. The rotation is recorded as a transformation matrix on the page rather than applied to pixels, so the renderer draws the original content at the corrected angle. Compare that with rotating an image in a photo editor, which resamples every pixel and softens the result.
Can different pages have different angles?
The angle applies to the whole document in one pass, because a stack fed through a scanner usually tilts consistently. For pages that differ, split the document, correct each part, and merge them again.
Should I deskew before running OCR?
Yes, and it makes a measurable difference. Character recognition segments a page into lines before it reads anything, and a tilt of more than about one degree starts merging lines together. Straightening first is the single cheapest accuracy improvement available.