OCR workflow: make scanned PDFs searchable
OCR (optical character recognition) adds searchable text to scanned pages. RedactNinja PDF recognises text in your browser: files are processed on your device and are not uploaded. The OCR language files themselves load from the website when you run OCR.
Recognise a scan
- Open a PDF and choose its OCR language in the side panel.
- Select Recognise scanned text for the current page, or Recognise document text for every page from that document.
- Check the recognised words. Correct mistakes and exclude words you do not want.
- If recognition failed on a page, exclude that page before you apply the other results. Pages that already contain text are kept unchanged.
- Apply the text, check the document and export it. Undo restores the previous state.
The languages available are English, French, German, Spanish, Polish and Welsh. Choose one language, or add a second, different language with Additional OCR language. Choosing a language does not translate the document.
OCR page layout offers Automatic, Single text block and Scattered text and labels. Choose Scattered text and labels for forms, diagrams or pages with short labels spread across the page.
Ordinary OCR does not add a second text layer to a page that already has text. That existing text may be incomplete, hidden or wrong, and any scanned areas on the same page are not recognised in this mode. Use one of the two options below instead.
Recognise a scanned area on a partly searchable page
- Select Recognise scanned area.
- Set the left, top, right and bottom edges as percentages of the page. Check the rectangle against the page preview.
- Keep the area clear of existing searchable text. If it overlaps existing text, the tool refuses rather than adding duplicate text.
- Select Recognise selected area, then correct and review the words, apply them and check the exported PDF.
This adds invisible searchable text to the area you chose and keeps the original page content. It does not replace existing incorrect text. Pages with rotated or right-to-left text may not pass the overlap check; use Rebuild document OCR for those.
Replace existing OCR with a rendered copy
- Choose the language, then select Rebuild document OCR.
- Read the limits and choose 150 or 300 dpi. A higher resolution takes longer and uses more memory.
- Select Rebuild and recognise text.
- Optionally correct the recognised words. Excluding a page leaves it without searchable text in the rebuilt copy; pages that failed must be excluded before you apply the other results.
- Apply the text layers and check the page appearance before export. Undo restores the original in the editor.
Rebuilding turns every page into an image first. Existing text layers, interactive forms, links, comments, bookmarks and attachments are not kept, and visible annotations may become part of the image. Text and drawings lose their sharpness when zoomed in. This is not redaction or archival conversion.
The original file is not changed. Signed PDFs cannot be rebuilt. If you have added annotations or redaction marks, export the document and open the exported copy before rebuilding.
Problems and limits
If a language file is missing or damaged, OCR stops with an error before recognition starts. Reload the page and try again.
Cancelling discards the proposed changes. The language settings are locked while OCR runs. OCR can miss text or produce words that look right but are wrong, and confidence values are not a guarantee. Check names, numbers and dates, especially in handwriting, complex layouts, faint scans and pages in more than one language.
For straightening, perspective correction and contrast before OCR, use Adjust scan image (the Clean up scans tool).