A scanned PDF looks like text but is really a picture of text — you can't search or copy from it. OCR (optical character recognition) reads the image and produces actual text. Here's how it works.
What OCR does
OCR analyzes the image of each page and recognizes the characters, outputting real text you can search, copy, and edit. It works best on clear, straight, reasonably high-resolution scans.
Languages
Good OCR supports many languages, including non-Latin scripts. Pick the language that matches your document for the best accuracy.
Privacy
Scanned documents are often personal. Browser-based OCR runs the recognition on your device, so the document isn't uploaded.
Step by step
- Open the OCR tool and add the scanned PDF.
- Pick the language of the document — accuracy drops sharply if this is wrong.
- Run the recognition. It happens in your browser, so the scan is not uploaded.
- Check the recognised text against the original before relying on it.
What usually goes wrong
OCR accuracy depends almost entirely on the scan
300 DPI, straight, evenly lit text recognises well. A phone photo taken at an angle in poor light will not, no matter which engine you use.
Handwriting is not what this does
OCR reads printed text. Handwritten notes come out as nonsense.
It is slower than you expect
Recognition is genuinely heavy computation, and doing it locally means your device does the work. A long document takes minutes.
Frequently asked questions
Is OCR always accurate?
Accuracy depends on scan quality. Clean, high-contrast, straight scans give the best results; faint or skewed scans produce more errors.
Can OCR read handwriting?
Most OCR is built for printed text. Handwriting recognition is far less reliable.
Why is my OCR result full of errors?
Almost always the source scan. Rescan at 300 DPI, straight and well-lit, and accuracy improves dramatically.
Can OCR read handwriting?
No. It recognises printed characters; handwritten text is a different problem.
Is my scanned document uploaded for OCR?
Not with a browser-based tool — the recognition runs on your own device, which matters given scans are usually IDs and certificates.