pc
Deskew · auto-detected

Deskew PDF — straighten tilted scans.

Phone or flatbed scans often arrive 1–3° off-axis. We detect the dominant text-line angle on each page and rotate it back to true horizontal — fractional, per-page, in your browser.

  • Stays in browser
  • No signup
  • Fractional · ±5°

Want to edit more on this PDF?

Sign, redact, watermark, autofill · stays in your browser

PDF
Drop a scanned PDF
or click to browse — stays in your browser
Detected · before
+0.0° straight
Corrected · after
+0.0° straight
Drop a PDF — we'll detect tilt angles automatically.
Mode
Background
Drop a scanned PDF
What it does

Per-page · sub-degree · auto-detected.

For each page, we render at low res, run a Hough-line detector on the rendered image to find the dominant text-line angle, then apply a counter-rotation when writing the output PDF. Sub-degree precision; the typical output lands within ±0.1° of true horizontal.

Hough-line detection

Standard CV technique for finding text baselines. Robust against handwriting and scanned forms.

Per-page correction

Each page detected independently — handles documents where pages were scanned individually.

Projection-profile fallback

When Hough confidence is low, we cross-check with a binary projection-variance method and pick the more reliable angle.

Background fill

Rotation creates corner gaps — fill with white (default) or leave original page background.

Skip already-straight pages

If a page is within ±0.2° we leave it alone — no needless re-rendering.

Vector-safe

We rotate by embedding each page as a form XObject, preserving text and font streams when the source is a vector PDF.

Four steps · seconds

How it works.

  1. Step 01

    Drop the PDF

    Drag a scanned PDF or click to browse.

  2. Step 02

    Detect angles

    We render each page and analyse text lines. Per-page tilt is shown in the summary.

  3. Step 03

    Apply

    Auto-detect picks the angle. Pages within ±0.2° are skipped.

  4. Step 04

    Download

    Straightened PDF saves locally. Original untouched.

Sensitive scans stay sensitive.

Deskewing is the kind of operation usually done on contracts, IDs, and medical forms — exactly the documents you don't want to upload. Ours runs in your browser, so the document never moves.

  • No file upload — verifiable in DevTools
  • No account, no email, no rate limit
  • No watermark on the output
  • Works offline after first visit
Common questions

About deskewing.

What's the difference between Deskew and Auto-rotate?

Auto-rotate fixes 90/180/270° flips (page is sideways or upside-down). Deskew fixes fractional tilt (page is 1–3° off true horizontal). They're often used together: auto-rotate first to get the page upright, then deskew to perfect the angle.

How accurate is the detection?

Typically within ±0.1° of true horizontal on documents with at least 5 lines of clear text. Single-image-per-page scans (no text) can't be deskewed automatically — coming v1.1 with manual mode to set the angle by eye.

What angles does it correct?

Up to ±5° in each direction. Beyond that, the page is probably misaligned at scan time and needs re-scanning. A manual ±15° slider is on the v1.1 roadmap if you really need extreme correction.

Will text get blurry from rotation?

For scanned PDFs (one image per page), there is a tiny resampling cost — well under 1 px shift at sub-degree angles. For text-based PDFs (rare for skew problems), the text is rotated as vectors with no blur.

What about handwriting scans?

Handwriting throws the line-detector off — pages with mostly handwriting often get misdetected. We flag them as low confidence and the manual mode (v1.1) will let you override per page.

Need to do more than this?

Open the full editor for signing, redaction, watermarks, autofill, and the rest of the toolbelt — same file, no upload, no signup.

Open the free editor