Accuracy depends on the source. Sharp, high-contrast typed text at 300 dpi recognizes very well; blurry scans and unusual fonts produce errors. Verify with the OCR Quality Checker.

Can OCR read handwriting?

No, not reliably. OCR targets printed text; handwriting should be transcribed manually.

Yes — free and unlimited, processed entirely in your browser.

Are my scans uploaded?

No. The tool uses locally bundled language data; your document never leaves your device.

What resolution should I scan at?

300 dpi is the sweet spot for most text. Higher resolution rarely improves accuracy further but makes files larger; lower resolution blurs characters and hurts recognition.

Can I OCR one page of a larger PDF?

The tool processes the whole document. If only one section needs OCR, split out those pages first with Split, OCR them, and merge the results back.

Does OCR fix blurry scans?

No — OCR reads what is visible. Blurry or low-contrast pages should be re-scanned; a better source always beats post-processing.

Is OCR the same as text extraction?

No. Text extraction reads the text layer that already exists in a digital PDF. OCR creates a text layer from images. A scanned PDF needs OCR before text extraction can return anything.

How long does OCR take?

Recognition runs locally, so speed depends on your device and the page count. Progress is shown while each page is analyzed, and you can cancel at any time.

Does OCR work on forms and tables?

It recognizes the printed text in them. Structured data like tables can then be extracted to CSV where the layout allows, using the Table Extraction tool.

Can I use the searchable result in other tools?

Yes — after OCR the PDF behaves like a digital document, so PDF to Word, Extract Text and other text tools all work on it as expected.

How to make a scanned PDF searchable with OCR: languages, accuracy and proofing

Q: Which languages are supported?

105+ languages are bundled locally with no CDN dependency, including English, Hindi, Spanish, French, German, Portuguese, Italian, Arabic, Russian, Chinese, Japanese, Korean, Bengali, and many more.

Guide · Updated 2026-08-13 · Extract cluster

OCR (optical character recognition) reads the text in scanned images and adds it to the PDF as a searchable text layer. You can then search, copy and extract the words even though the original is a picture. Recognition quality depends on image sharpness and the language of the document; the Paperly tool runs in your browser with locally bundled language data, so files are never uploaded.

A scanned PDF is a photo of a document. You can read it, but the computer cannot — there is no text layer, so search, copy and extraction all come up empty. OCR solves that by analyzing the image, recognizing the characters, and writing the recognized text into the file.

This guide explains how OCR works, what affects accuracy, which languages are available, and how to get the best results from the Paperly OCR tool — which processes everything locally in your browser.

What OCR adds to a PDF

OCR does not replace your pages. It adds a hidden text layer on top of each scanned page, containing the recognized words positioned to match the image. The page still looks exactly the same — but now the words are searchable, selectable and extractable. You can search a term in the reader, copy a sentence, or run a text-extraction tool on the result.

What affects recognition quality

Image sharpness — blurry scans produce blurry recognition. Scan at 300 dpi or higher for the best results.
Contrast — clean black-on-white text recognizes far better than grey-on-grey or photos.
Language — the recognizer uses language-specific data; picking the right language matters.
Fonts and layout — unusual fonts and heavy decorative styling lower accuracy.

Supported languages

The Paperly OCR tool bundles language data for 105+ languages including English, Hindi, Spanish, French, German, Portuguese, Italian, Arabic, Russian, Chinese, Japanese, Korean, Bengali, and many more. The language data is served from the site itself — there is no external CDN and no upload. You can pick one language or several: selecting multiple helps mixed-language documents (for example English + Hindi), at a small cost to recognition speed. For the fastest single-language run, pick just the one language you need.

How to OCR a PDF step by step

Open the OCR PDF tool.
Upload a scanned PDF.
Choose the document language — pick one for speed, or several for mixed-language documents.
Process — each page is analyzed and the text layer is added.
Download the searchable PDF.
Verify by searching for a known word or running OCR Quality Check.

Verifying OCR results

OCR is not perfect. Handwriting, low-quality scans and unusual layouts will contain errors. The OCR Quality Checker tool reports a confidence estimate per page, and you can always spot-check by searching for words you know are in the document. When a word fails to recognize, re-scanning at higher resolution is the most reliable fix.

What OCR is not

OCR recognizes text in images; it does not understand meaning, layout or handwriting reliably. It will not fix a blurry scan, it cannot interpret complex tables as structured data, and it treats handwritten notes as unrecognizable marks. For structured data, use the Table Extraction tool after OCR; for meaning, a human still reads.

Privacy

All recognition runs locally in your browser using bundled language data. The images never leave your device, which matters for sensitive scans — medical records, contracts and personal documents are processed without any server ever seeing them.

The language data (trained models for all seven supported languages) is served from the site itself, not from a third-party CDN. That keeps the privacy story simple: your document travels nowhere, and the software that reads it is served by the same site you are using.

What OCR works best on

Clean typed documents — letters, invoices, reports, forms printed on white paper.
Standard fonts — serif and sans-serif text at normal sizes.
Single-language documents — one run handles one language best.
Straight, well-aligned pages — skew hurts recognition more than minor noise.

If your scans are heavily skewed or low-resolution, straighten and re-scan them first; OCR quality improves more from a clean source than from any setting in the tool.

After OCR: what to do with the searchable PDF

Once a scan has a text layer, the whole Paperly workflow opens up. Extract Text pulls the words for analysis, PDF to Word converts the recognized text into an editable document, and Table Extraction can structure the recognized content into CSV where the layout allows. Search inside the PDF also starts working — in your reader and in the browser-based tools.

OCR is the bridge between paper and data. The recognized text layer is what makes every downstream text tool work on a document that was originally just a picture.

OCR and searchable archives

The most common use of OCR is archiving. A folder of scanned contracts, invoices or medical records is useless for retrieval until the text inside it can be searched. Adding a text layer turns that folder into a searchable archive: every document can be found by any word it contains, without opening files one by one.

For archival-quality output, run the searchable PDF through the PDF/A tool afterwards to normalize the document for long-term storage. Combined, the two steps produce a searchable, standards-shaped archive copy that preserves both the original look and the newly recognized text.

Frequently asked questions

No — the visual page is untouched. OCR only adds a hidden, searchable text layer on top of the existing image.