FREE
Make PDF Searchable
Add a hidden text layer to scanned PDFs so you can search, select and copy text.
Launch
Convert & Export
OCR PDF — Recognise Text in Scans
Scanned documents are just pictures of text. OCR (optical character recognition) reads the words and adds an invisible text layer, so you can search, select and copy text — or export it as plain text. Recognition runs on your device.
- Runs in your browser — no upload
- Free — no sign-up
How to use OCR PDF
- 1Open a scanned PDF, or photos/scans of pages (JPG, PNG, WebP).
- 2Choose the document language(s) — up to three.
- 3Run OCR, check the confidence per page, then download the searchable PDF or the text.
Why use OCR PDF on PDF NOVA
Searchable PDFs
Each page keeps its original look; an invisible text layer is added on top, so the PDF becomes searchable and copyable.
Six languages
English, French, Arabic, Spanish, German and Italian, including accents and right-to-left Arabic.
Files stay on your device
Processing happens in your browser. Your document is never uploaded to a server, so there is nothing to delete afterwards.
Processing on your device
OCR PDF processes your file in your browser on your device. The file is not uploaded for processing. Download the result to keep a permanent copy. Your active PDF is saved locally in this browser for recovery across refreshes. Close Document removes it; recovery expires after 24 hours and is cleaned up when PDF NOVA next opens. Unsaved changes are not restored.
Related PDF tools
- Launch
- Launch
- Launch
Frequently asked questions
Are my files uploaded anywhere?
No. This tool runs entirely in your web browser using JavaScript and Web Workers. The file is read from your device, processed in memory and saved back to your device.
How accurate is OCR?
Clean, upright scans at 300 DPI usually recognise very accurately. Low-resolution photos, handwriting, skewed or sideways pages and decorative fonts reduce accuracy — words with low confidence are highlighted so you can check them.
Why is the first run slower?
The recognition engine and the language model (1–3 MB per language) are downloaded once, then cached by your browser.