Make a Scanned PDF Searchable
How to OCR a PDF
A PDF from a scanner, a copier or a phone scanning app, up to 100 MB.
Pick the language of the document, so that special letters are recognised correctly.
A PDF that looks identical but can be searched, or a text file with just the words.
How to tell that a PDF needs OCR
Open the PDF and try to select a word with the mouse, or search for a word you can see on the page. If nothing can be selected and the search finds nothing, the pages are pictures. That is what scanners, copiers and phone scanning apps produce unless text recognition is switched on.
What changes in the file
- The pages look the same. The scanned picture is kept; the recognised words are placed behind it, invisible, each at the position where it stands on the page.
- Search, select and copy work. Ctrl+F finds words, text can be marked and copied, and screen readers can read the document aloud.
- Pages that already contain text are left alone. In a PDF that mixes typed and scanned pages, only the scanned ones are processed.
- The file size stays close to the original. In our test, a one-page scan of 84 KB became 44 KB, because the picture was also stored more efficiently.
How accurate is it
Recognition is done by Tesseract through OCRmyPDF, both open-source. A clean scan at 300 dpi of printed text is read almost without errors. Skewed pages, coffee stains, stamps over text, very small print and faxes lower the accuracy; handwriting is not recognised. The hidden text is only as good as the recognition, so a search can miss a word that was misread.
The language setting is the most important choice. With the wrong language, letters such as š, đ, ž, ü or ç come out as plain s, d, z, u and c, and searching for the correct spelling then finds nothing. "Mixed or not sure" tries all languages at once and is noticeably less accurate for these letters.
Searchable PDF or text file?
Choosing the result
| Result | What you get | Use it for |
|---|---|---|
| Searchable PDF | The original pages with a hidden text layer | Archiving, sending on, searching in contracts and invoices |
| Text file | The words only, page after page, in UTF-8 | Pasting into another document, translating, analysing |
For an editable Word document with headings and paragraphs, use PDF to Word; it runs the same recognition automatically when it finds a scan. Details are in Scanned PDF to Word with OCR.
Limits
- Processing stops after 5 minutes. A scan of a few dozen pages is fine; for very long documents, divide the file with PDF Split and process the parts.
- Password-protected PDFs are refused, and so are digitally signed ones, because changing the file would break the signature.
- A single photo or screenshot goes through Image to Text. Pages that are upside down or sideways should be turned first with Rotate PDF.
- The file is uploaded over HTTPS, processed on our server and deleted within 4 hours.
Frequently Asked Questions
What does OCR do to a PDF?
It reads the text in the scanned pages and adds it as an invisible layer. The PDF looks the same but can be searched, selected and copied.
Will my PDF look different afterwards?
No. The scanned pages are kept as they are. Only the hidden text is added.
Which languages can it read?
English, Bosnian, Croatian, Serbian (Latin and Cyrillic), German, French, Italian, Spanish, Russian and Turkish.
My PDF already has some text. What happens?
Pages that already contain text are left unchanged. Only pages without text are processed.
Can I get just the text?
Yes. Choose "Text file" as the result to download the recognised words as a .txt file.
How long is my file kept?
The upload and the result are deleted from the server within 4 hours.