PDFJadoo › Scan & OCR › Deskew PDF
The PDFJadoo genie with his golden lamp, rising in a swirl of gold magic

Straighten crooked scans automatically

A scanner fed slightly sideways leaves every page tilted. This measures each page's tilt by analysing its text lines, rotates it level, and reports the angle it fixed on every page.

Runs 100% in your browserFiles never uploadFree · no sign-up
Drop your scanned PDF here

or click to choose a .pdf file

Honest note: pages are rebuilt as high-quality images at 200 DPI (straightening a scan is inherently pixel work — the scan was pixels to begin with). Pages already level are copied untouched. Works on tilts up to about ±15°; rotate grossly turned pages first with Rotate PDF.

🔒 Private by design: this tool runs entirely in your browser. Your file is never uploaded, never stored, and works even with the internet off.

How to deskew a PDF

  1. Drop your scanned PDF into the circle.
  2. Leave the strength on Auto for normal scans.
  3. Press Straighten the pages.
  4. Read the per-page angle report.
  5. Download the levelled PDF.

The deskew engine looks for the angle at which the page's rows of text line up most strongly (a projection-profile search), then rotates by exactly that much with high-quality resampling. Already-straight pages pass through untouched.

The measurement is a small piece of signal processing. Each page is rendered as a low-resolution probe and the engine sweeps candidate angles, projecting the page's ink onto the horizontal axis at each one. When the angle is right, text rows stack into sharp peaks; when it is wrong, they smear. The peak wins, and the page is rotated by exactly that correction at a full 200 DPI rendering, then placed back on a page of the original size.

The two strengths are thresholds, not amounts of force. Auto ignores anything under 0.3°, which is below what most eyes notice, and copies those pages through completely untouched — no re-rendering, no quality change. “Every page” lowers the bar to 0.1° for perfectionists and print work. Either way the status line reports the measured angle for every page, so you can see what was fixed and what was left alone.

Why scans tilt — and the one thing deskew cannot fix

Scanners tilt pages for unglamorous reasons: a sheet that enters the feeder a fraction crooked, a page that curls against the glass, a book spine that will not let the page lie flat. The result is a document that looks faintly drunk and — less obviously — OCR software reads it measurably worse, because text-recognition expects level lines. Deskewing is the unglamorous fix that makes scans look professional and machine-readable again.

Know its boundary, though. Deskew corrects one thing: a uniform rotation of the whole page. It does not undo perspective distortion (the trapezoid look of a page photographed at an angle), wavy lines from a scanner that stuttered mid-pass, or a page that was cropped mid-sentence. Those need a re-scan, not a rotation. Within its boundary — the everyday feeder tilt of a degree or three — the fix is exact, measured per page rather than guessed for the whole file.

Related wishes

Or browse the full deck of 130 free PDF tools.

Deskew PDF — fair questions

Will text PDFs work?

Deskew targets scanned pages. Born-digital pages are detected as level and copied unchanged.

What tilt range is covered?

About ±15°, which covers every realistic scanner feed error.

Does quality drop?

Pages that need rotation are re-rendered at 200 DPI — indistinguishable for scanned material. Untouched pages lose nothing.

My pages are phone photos and look trapezoid-shaped — will this fix them?

Only partly, and it is worth knowing the boundary. Deskew corrects a page that is uniformly rotated — the whole page tipped by the same angle. A photographed page usually suffers perspective: the far edge looks smaller, lines converge, and no single rotation undoes that. Re-shoot the page square-on (the Scan tools here help), then deskew what remains.

Should I deskew before running OCR?

Yes, in that order. OCR engines read along assumed text lines, and even a two-degree tilt measurably costs accuracy. Straighten first (Rotate PDF for sideways pages, then Deskew for tilts), then run OCR PDF on the levelled result for the cleanest text layer.

Why does a straightened page look very slightly softer?

A page that needed rotation is rebuilt as a 200 DPI image, so detail far above 200 DPI in the original scan is resampled down. For ordinary office scanning (200–300 DPI) the difference is negligible in print and on screen. Pages measured as level are copied untouched and lose nothing at all.

Back to All Tools