When to deskew, clean or force OCR

Deskew a tilted scan

Deskew straightens small rotations in the page image. Use it when printed baselines slope. It is different from rotating a sideways page by 90 degrees. Inspect edges and handwritten marks after processing.

Clean background noise

Cleaning can reduce light background noise before recognition. It does not restore letters that are blurred, cut off or absent. Faint annotations and photographs need review. Keep the original.

Force recognition when text copying is unusable

Forced OCR renders every page into an image, then recognizes a new text layer. Use it when ordinary extraction pastes symbols or existing recognition misses visible content. It flattens interactive content and changes the PDF structure. It does not retain digital-signature validity or certify accessibility.

Keep automatic OCR for usable mixed documents

When digital text copies correctly, automatic OCR can keep it while recognizing scanned pages. Compare the owned mixed example with its recognized result.

Check the output

1. Choose languages actually printed on the page.
2. Process a representative excerpt with the appropriate settings.
3. Search a visible phrase and copy it into a text editor.
4. Check small print, numbers, columns, page edges and form content.

OCR is fallible. A successful download does not mean every word is correct.

Related workflows

Fix text that copies as symbols · Scanned PDF to TXT · OCR a large PDF