How to OCR a scanned PDF privately, on your device
A scanned PDF is really just images — you can't search or copy its text. OCR fixes that by adding a text layer. But uploading a sensitive scan to a cloud OCR service isn't ideal. Here's how to OCR a PDF entirely on your own device.
- Open an on-device OCR tool. Open the free OCR PDF tool on taimurtools.com. Recognition runs in your browser, so the scan is never uploaded.
- Add your scanned PDF. Select the scanned PDF. It stays on your device throughout.
- Run OCR. The tool recognises text on each page locally and adds an invisible, selectable text layer.
- Download the searchable PDF. Download the result. You can now search, select, and copy text from what used to be a flat image.
What OCR does to a scanned PDF
A scanner produces a picture of each page. To you it looks like text; to the computer it is only pixels, which is why you can't search it, select a sentence or copy a figure. OCR (optical character recognition) reads those pixels and works out which letters they show. The OCR PDF tool then adds the recognised text as an invisible layer behind the page image. The page looks exactly as before, but now you can search it, select text and copy it into another document. It supports 12 languages, and the whole job runs in your browser using the Tesseract.js recognition engine.
When it's worth doing privately
- Contracts and legal documents you need to quote from or search across.
- Financial and medical paperwork (statements, invoices, reports) that shouldn't sit on a third-party server.
- ID documents and HR files, where a leak would be a real problem.
- Archives of old scans you want to make searchable without paying per page.
Getting a good result
- Scan straight and sharp. Skewed, blurry or very low-resolution scans recognise worse; if you can, re-scan at a normal document setting.
- One language at a time works best. The tool supports many languages, but a document that mixes several will recognise less reliably.
- Printed text beats handwriting. OCR is built for typed text; handwritten notes recognise poorly or not at all.
- Check it. Open the result and search for a word you can see on the page. If it's found, the text layer is in place.
After OCR
The searchable PDF behaves like any other: you can merge it with other documents, compress it if the scan is large, or split out the pages you need. For a single photo or image of text rather than a PDF, use Image to Text instead.
How the tool works, step by step
OCR PDF does three things to each page in turn. It draws the page as an image with PDF.js. It reads that image with Tesseract.js, an OCR engine that runs inside your browser tab. Then it writes the recognised text back into the page as an invisible layer with pdf-lib. The visible page is never altered, so a misread word can never spoil how the document looks; at worst, a search will not find it.
Because your own device does the recognition, it is slower than a cloud service with dedicated servers. Expect a long scan to take a while, and keep the tab open until it finishes. On first use the OCR engine and the language data are downloaded to your browser; that download is the engine, not your document.
Choosing the language
Pick the language the document is written in before you start. The engine uses a separate model for each language, and reading a French letter with the English model produces many errors. If a document mixes two languages, run it with the main one; the other will recognise less reliably.
What accuracy to expect
Accuracy depends almost entirely on the scan. Clean, straight, high-contrast typed text recognises well. Faint photocopies, skewed pages, unusual fonts, tables with ruled lines, stamps over text and busy backgrounds all produce errors. Handwriting is not supported. OCR also confuses characters that look alike, such as the letter O and the digit 0, or I, l and 1, so check account numbers, amounts and dates against the page before you rely on the text.
OCR PDF or Image to Text?
Use OCR PDF when you have a PDF and want to keep it as a PDF that you can search. Use Image to Text when you have a single picture, such as a screenshot or a phone photo, and just want the words to paste somewhere. Image to Text reads 13 languages, including Urdu, Arabic and Hindi, handles up to 20 images in one run and lets you correct the text before copying; use OCR PDF when you need a searchable PDF rather than plain text.
Limits
Free use is capped at five runs per day for this tool, counted in your browser and reset at midnight UTC. There is no file-size cap, but a long, high-resolution scan uses a lot of memory and can stall a phone; split it with Split PDF and process the parts.
FAQ
Does OCR change how my PDF looks?
No. It adds an invisible text layer to the existing page, so the document looks identical but becomes searchable.
Is my scan uploaded for OCR?
No. Recognition runs in your browser with Tesseract.js. The engine and language data are downloaded to you; your document is not sent anywhere.
Which languages are supported?
12. Choose the document's language before you start, because each language uses its own recognition model.
Can it read handwriting?
No. It is built for printed text.
Why is it slow?
Your own device renders and reads every page. Long or high-resolution scans take time; split them into parts if a phone struggles.