Running OCR on a PDF
Choose the scanned PDF
Hundreds of pages are fine. The page grid shows each page; those found to have text already are labelled has text.
Set the language and resolution
Pick English, Welsh or both. 300 dpi suits most scans; 200 dpi is faster and 400 dpi helps with small print.
Choose the workers
Workers read pages in parallel, and each uses about 700 MB of memory. Leave the default on a modest laptop; raise it on a well-equipped desktop.
Read the pages
Press Find text pages and read the rest. A progress bar estimates the time left, and you can pause, resume or stop at any point.
Save and check
Save the searchable copy, or download just the recognised text as a .txt file. Search the saved PDF for a word you can see on a page.
Example: a mixed file of typed and scanned pages
A letting agent combines a typed tenancy summary (pages 1 to 3, already text) with a scanned inventory of 22 pages, then needs to search the inventory for every mention of damp. Only the scanned pages need reading.
After saving, the tool reopens the new file and looks for the recognised words on each page. A high match rate confirms the hidden layer was written correctly; a low one is a cue to check those pages by eye.
tenancy-pack.pdf, 25 pages Language: English, 300 dpi, 2 workers, Only these pages: (blank)
Pages 1-3 skipped (has text) Read 22 pages, skipped 3 pages that already had text Saved tenancy-pack-searchable.pdf with a hidden text layer
Honest limits
OCR reads clear printed text well. Handwriting, rubber stamps, faint carbon copies and skewed or low-resolution scans may be missed or misread. The hidden layer is for finding and copying; if you quote a figure or a clause, check it against the page image.
The languages offered are English, Welsh, and English with Welsh together. If a page fails to read, the grid marks it and the rest of the file is still saved.
Questions people ask
How do I make a scanned PDF searchable for free?
Open it in this tool, choose the document language, press Find text pages and read the rest and then save the searchable copy. The recognition engine runs in your browser, so there is no upload, account or page limit, although long files take a while on slower machines.
Will OCR change how my PDF looks?
No. The scanned page images stay exactly as they were; only an invisible text layer is added on top, aligned with the printed words. When you search, your PDF reader highlights the word on the image, and selecting text copies the recognised characters.
How accurate is the OCR?
Clean, typed documents scanned at 300 dpi are usually read very well. Accuracy drops with handwriting, small or decorative fonts, stamps, poor contrast and crooked pages. After saving, the tool reports how many recognised words it found again in the new file, and you can view the recognised words outlined on each page.
Why does it need an internet connection the first time?
The open-source Tesseract recognition engine and its language data, about 6 MB for English, are downloaded from a public code library the first time you use it. Your PDF itself is not sent with that request; it is read on your computer by the downloaded engine.
Where your data goes
On first use your browser downloads the open-source Tesseract engine and language data from cdn.jsdelivr.net, a public CDN. Your PDF is never uploaded: pages are drawn and read by that engine on your own computer, and the searchable copy is created locally.
Checked against the tool on