OCR PDF

Runs in browser OCR & Text

Turn a scan into a PDF you can search, select and copy from. Each page is recognised locally with Tesseract and the words are written back as invisible text exactly where they are printed; the page image is untouched. Pages that already have text are skipped.

Privacy: Processed locally in your browser. Your file is not uploaded.

What is OCR?

OCR (optical character recognition) turns the pixels of a scanned page or photo into text: it finds lines and words, recognises characters with a model trained per script, and produces text you can search, copy and index. Accuracy depends on resolution (300 dpi is the usual target), contrast, skew and the script — Arabic and Devanagari need their own language models — and the output should be proofread where it matters.

A page without a text layer is recognised, and each confident word is written as invisible text over the scan, so the page looks the same but can be searched.
Question 1: Does the page already have a text layer?Asked for every selected page

  • Answer: Yes

    Question 2: Is “Skip pages that already have text” ticked?

    • Answer: Ticked (default)

      Outcome: Left as it is: no OCRIts text can already be selected and searched

    • Answer: Unticked

      Outcome: Recognised anyway, like a scan

  • Answer: No: a scan, only pixels

    Question 3: Is the recognised word at least 30% confident?The page is rendered at 150, 200 or 300 dpi and read by Tesseract in your browser

    • Answer: Yes

      Outcome: Written as invisible text over its printed wordThe scan is unchanged; you can now search, select and copySearchable PDF

    • Answer: No

      Outcome: Left out of the text layerStill kept in the text preview and .txt download

How to use OCR PDF

  1. Drop the scanned PDF.
  2. Pick the language(s) of the text and the resolution.
  3. Click "Make searchable".
  4. Download the searchable PDF (and the text, if useful).

OCR PDF features

  • Tesseract OCR in the browser (WebAssembly), models served from this site
  • English, Arabic, Hindi — alone or combined
  • Invisible text layer placed word by word over the scan
  • Pages with text already are skipped
  • 150 / 200 / 300 dpi
  • Confidence shown; low-confidence words left out of the text layer
  • Text preview and .txt download too

OCR PDF example

A 20-page scanned contract in Arabic and English

Input:

contract-scan.pdf, language: Arabic + English, 200 dpi

Output:

contract-scan-searchable.pdf — same pages, now selectable and searchable; average confidence 91%

Frequently asked questions about OCR PDF

Does the OCR happen on a server?

No. Tesseract runs as WebAssembly inside your browser and the language models are downloaded from this site once (about 3 MB per language). The scan never leaves your device.

What does "searchable" mean here?

The recognised words are added as an invisible text layer over each page, sized and placed to match the printed words. The page still looks like the scan, but you can search it, select text and copy it, and full-text search in file managers finds it.

How accurate is it?

On a clean 200–300 dpi scan of printed text, typically well above 95% for Latin script and somewhat lower for Arabic and Devanagari. The result shows the average confidence, and words below 30% confidence are left out of the invisible text layer rather than written as garbage; the text preview and .txt download keep every recognised word. Handwriting is not recognised.

Which languages are supported?

English, Arabic and Hindi, alone or combined (for example Arabic + English for a bilingual contract). Choosing the right language matters more than any other setting.

Will it slow down or grow my file?

The text layer adds a few kilobytes per page. Recognition takes roughly 2–6 seconds per page on a laptop, longer on a phone.

Is my PDF uploaded to a server?

No. The file is opened and processed by your browser with WebAssembly and JavaScript; nothing is sent to Mutqan Soft or anyone else. Close the tab and the file is gone. Downloading the result needs a free account — no subscription, no payment.

What are the size limits?

Up to 250 MB and 2,000 pages per operation on a desktop browser, 60 MB and 500 pages on a phone, because everything runs in your device’s memory. The tool tells you before it starts if a file is over the limit.

Can I use it on a password-protected PDF?

Yes, if you know the password: the file list asks for it once and every operation then works on the decrypted pages. The output is saved without a password unless you protect it again.

Technical notes

Each page is rendered by pdf.js at the chosen dpi (rotation applied) and passed to Tesseract (tesseract.js 7, LSTM `best_int` models). The words and their pixel boxes are mapped back through the page’s viewport transform and written as invisible text (render mode 3) with the embedded Cairo / Noto Sans Devanagari faces, each word stretched to its box with a text matrix, so selection follows the print. The scan’s content stream is untouched.

References