Start with the page in front of you.
Open the PDF and choose OCR. Under Pages to recognize, Current page checks the page you are viewing. The hint shows its page number. Choose All pages for the whole document or Page range and enter numbers such as 1-3, 5. Repeated page numbers are processed once.
Use an installed recognition language.
Choose Recognition language, then Make searchable. Recognition runs through Windows on this PC. Pagekind does not upload the PDF or download language packs. If no OCR language is installed, install one in Windows language settings and reopen the document. Each open PDF tab keeps its page and language choices while the window stays open.
See what changed.
The result reports how many pages received searchable text and how many already contained text. Cropped page boundaries and the text angle reported by Windows are used to place the invisible layer. The visible scan, page rotation and existing annotation positions stay in place. Choose Find text in this PDF, search for a name or number, and inspect the result.
If every selected page already contained text, Pagekind reports No new Undo step was added. Existing text is kept. A page with a small text header is also skipped, even if the rest is an image. Replacing an existing OCR layer remains separate work. To correct recognition before applying it, start from the scan and use the word review below.
Recognition finds words, not the original typeface.
Pagekind uses recognized words and their positions to add an invisible text layer. It does not identify the exact font, recreate the original typeface or change the scan’s printed letters. The visible font stays part of the original page image. Correcting a word in Review words on this page changes the searchable text, not that image.
For clearer results, choose the language used on the page and start with a sharp, upright scan. Check names, dates and numbers against the original, then try Find and selecting and copying a passage. Unusual lettering, handwriting, blur and complex layouts can reduce recognition accuracy.
Keep control of the result.
Use Undo to remove an applied recognition step and Redo to restore it. Changes stay in the working document until you choose Save a copy. Your source PDF remains unchanged. An invalid page range stays available to correct, and a lost application reply is checked against that operation before another change proceeds.
Review a page before applying its words.
Choose Review words on this page to compare the original scan with its recognized words. Select a word to see its outline, use the scan zoom for a closer look, and type a correction. Search the word list to find a name or number quickly. Use Next word or Enter to move through the list.
Choose Use original to reset a word or Omit word to leave it out of the searchable text. Corrections are kept locally while you type. Wait for Corrections kept locally, then choose Apply corrections. Applying changes the searchable text and adds an Undo step; it does not change the printed appearance of the scan. Use Find to check the result, then Save a copy.
Come back to a kept review.
Choose Keep for later. When you return, open Recover and choose Continue word review. Pagekind opens the kept scan in its own tab with your corrections. An encrypted PDF requires its password again after restarting. If an opening or edit was interrupted, Pagekind checks its earlier result before retrying.
Completed reviews are available through Manage completed review. You can discard a kept review when it is no longer needed; your PDF remains. If a discarded scan file could not be removed, Recover offers a cleanup retry. Read more about recovering local work.
A second look at small print.
When the normal recognition pass finds no words, Pagekind can automatically try the page once at higher detail, within its image-size limits. This also applies when you open a word review. It can help with some small printed text; it does not guarantee that every typeface or page will be recognized.
A page with any recognized words keeps that first result. The extra pass does not repair partially recognized text or replace an existing text layer, and blank pages can take longer to finish.
Check important words.
If no words are recognized, Pagekind keeps the PDF unchanged and asks for a clearer scan. In a mixed selection, pages with no recognized words stay unchanged and are reported separately from pages that received searchable text.
OCR can mistake names, dates and numbers. A clearer scan and the right language can help. Words at different angles on the same scan may be recognized only partially. Word review covers one scanned page at a time and supports bounded corrections with available font coverage. It does not provide region recognition, replacement of existing text, complex-script shaping, recognition confidence scores, semantic reading order, handwriting guarantees or automatic accessibility tagging.
PDFs with signature, certification or timestamp structures need an unsigned original before OCR changes their contents. This structural check does not verify signer identity or certificate trust. Password-protected copies retain their encryption when recognition succeeds.
Included in the 0.65.0 Windows engineering preview. See the privacy note for local processing and recovery information.