Recognize the text of a PDF (OCR)
A scan is only an image: nothing in it can be searched, nothing copied. Character recognition adds a text layer under the image — the look does not change, but the document becomes searchable.
Free · 100% local — nothing is uploaded- 1Open your scanned PDF.
- 2Start the recognition and choose the language.
- 3Download the PDF, now searchable.
How to recognize the text of a scanned PDF
A scanned document is only a photograph of a page: nothing in it can be searched, nothing copied, and a search engine sees nothing but gray. Optical character recognition reads the image and lays an invisible text layer underneath: the look of the scan does not change by a pixel, but the document becomes searchable, selectable and copyable, and it can then be converted to Word or to Excel. Open the PDF, choose the document’s language, start the recognition, and download. What is rare here: everything is computed inside your browser, the engine included — a medical file, a payslip or a notarial deed does not have to be entrusted to an online service in order to become readable. Recognition takes a few seconds per page, and depends on how sharp the scan is.
Which languages, and how reliable
Six languages are carried, English and French among them with their accents and ligatures; they are downloaded once, on first use, then kept for the next times. No recognition is a hundred per cent reliable: a clean, straight page scanned at 300 dots per inch gives an almost perfect text; a photo taken askew, handwriting or a stamp across the text give errors. The recognized text can be read back and corrected afterwards in the editor, like any other text.
Frequently asked questions
Is the OCR done on a server?
No — which is rare, and is the whole point: the recognition engine runs inside your browser. A confidential scanned document goes nowhere.
Which languages are recognized?
French, English, Spanish, German, Italian and Portuguese — plus a French + English mode for mixed documents. The model for the chosen language downloads on first use, then stays on the device.
Is the look of the scan changed?
No. The image stays as it is; the recognized text is laid underneath, invisible, purely for searching and copying.
How long does recognition take?
A few seconds per page on a recent computer, longer on a phone: all the computing happens on your machine, with no queue and no quota.
Is the recognized text a hundred per cent reliable?
No, and no engine is: a clean, straight scan gives excellent results, a tilted or screened photocopy far less. Recognized text is there to search and to copy; proofread it before using it as a source.