How does automatic detection work, and how accurate is it?

Each page is turned into black and white, and for every candidate angle the ink pixels are projected along that direction. When the angle matches the text lines, the projection forms sharp peaks (lines) and valleys (line spacing), giving the highest concentration. The search runs from −15° to +15° in 0.5° steps, then refines around the best value in 0.1° steps. Text-heavy pages are usually accurate to 0.1°; pages that are mostly pictures may show “No text lines found” and stay at 0°, in which case you can type the angle yourself.

Will the text still be searchable after deskewing?

No. Straightening rotates each page as an image, so every output page is a JPG and any text layer in the original PDF is rasterized. Scans are images anyway, so little is lost; if you need search, run the straightened file through our Scanned PDF OCR tool — recognition usually improves once pages are level.

When should I use auto-fix for sideways or upside-down pages?

Use it when some pages were scanned turned around or sideways. The tool compares horizontal and vertical projections to see whether text lines run across or down the page, then uses the fact that Latin text has more ascenders (b, d, h, k…) than descenders (g, p, y…) to tell upright from upside down. It suits horizontal English or mixed text; vertical CJK or picture-only pages may be misjudged, so it’s off by default and every page’s rotation can be changed back by hand.

Will the file get bigger? Do page sizes change?

Page sizes stay exactly the same (width and height swap only when you rotate a page by 90° or 270°). File size depends on resolution and JPG quality: 200 DPI at 85% is usually similar to a typical scan; choose 150 DPI or lower quality for smaller files, or 300 DPI for print and archiving. Everything runs in your browser and nothing is uploaded.

📠

📠Deskew PDF Online - Straighten Crooked Scans Automatically

When paper goes into a scanner or copier at an angle, every page of the PDF comes out tilted — awkward to read, ugly to print, and worse for OCR. This tool renders each page, binarizes it and measures the angle of the text lines with a projection profile: it searches ±15° coarsely, then refines to 0.1°, and rotates the page back, filling the corners with white. After detection you get a before/after preview of every page with horizontal guide lines, and you can type a different angle or rotate a whole page by 90°, 180° or 270°. Optionally paint dark scanner borders around the edges white, and auto-fix sideways or upside-down pages (judged from text-line direction and the balance of ascenders and descenders, which suits horizontal text). Besides PDFs you can upload several scanned images (JPG, PNG and more) to get one straightened PDF. Output pages keep the original page sizes, with 150, 200 or 300 DPI and adjustable JPG quality. Everything runs in your browser; files are never uploaded.

Free · No sign-up · Runs in your browserLast updated: 2026-10-26

Loading tool...

How to use Deskew PDF

1Upload a scanned PDF or scan images (several allowed)
2Choose resolution, JPG quality, border removal and auto-rotation, then click Detect skew
3Review each page’s before/after preview, adjust angles or rotation if needed, then click Create straightened PDF and download

FAQ

How does automatic detection work, and how accurate is it?: Each page is turned into black and white, and for every candidate angle the ink pixels are projected along that direction. When the angle matches the text lines, the projection forms sharp peaks (lines) and valleys (line spacing), giving the highest concentration. The search runs from −15° to +15° in 0.5° steps, then refines around the best value in 0.1° steps. Text-heavy pages are usually accurate to 0.1°; pages that are mostly pictures may show “No text lines found” and stay at 0°, in which case you can type the angle yourself.
Will the text still be searchable after deskewing?: No. Straightening rotates each page as an image, so every output page is a JPG and any text layer in the original PDF is rasterized. Scans are images anyway, so little is lost; if you need search, run the straightened file through our Scanned PDF OCR tool — recognition usually improves once pages are level.
When should I use auto-fix for sideways or upside-down pages?: Use it when some pages were scanned turned around or sideways. The tool compares horizontal and vertical projections to see whether text lines run across or down the page, then uses the fact that Latin text has more ascenders (b, d, h, k…) than descenders (g, p, y…) to tell upright from upside down. It suits horizontal English or mixed text; vertical CJK or picture-only pages may be misjudged, so it’s off by default and every page’s rotation can be changed back by hand.
Will the file get bigger? Do page sizes change?: Page sizes stay exactly the same (width and height swap only when you rotate a page by 90° or 270°). File size depends on resolution and JPG quality: 200 DPI at 85% is usually similar to a typical scan; choose 150 DPI or lower quality for smaller files, or 300 DPI for print and archiving. Everything runs in your browser and nothing is uploaded.