OCR PDF: make scanned PDFs searchable

A scanned PDF is a stack of pictures. You can read it, but your computer cannot: search finds nothing, you cannot copy a sentence, and screen readers have nothing to read. Optical character recognition (OCR) looks at each page image, finds the letters and words, and turns them into real text.

The tool offers three outputs:

The searchable PDF looks exactly like your scan, because the page images are kept. The text layer sits underneath, lined up with the words, so search, select and copy work as in a normal PDF.

Choose the language of the document. 30 languages are available: English, Spanish, French, German, Portuguese, Italian, Dutch, Polish, Czech, Romanian, Swedish, Hungarian, Turkish, Russian, Ukrainian, Greek, Arabic, Hebrew, Persian, Urdu, Hindi, Bengali, Tamil, Thai, Vietnamese, Indonesian, Chinese (Simplified and Traditional), Japanese and Korean, plus English combined with Spanish, French or German for bilingual documents. The right language matters, because recognition uses a dictionary and letter shapes for that language.

Each language is downloaded only when you pick it, and the size is shown next to its name (0.4 to 3 MB). Your browser keeps it, so the second run starts straight away. The searchable PDF keeps every character, including Chinese, Arabic, Devanagari and Thai, because the hidden text layer does not depend on a font for that script.

Accuracy depends mostly on the scan:

• Resolution: 300 dpi is the sweet spot. Below 200 dpi, small text gets unreliable.
• Straight pages: skewed or curved pages (a thick book) reduce accuracy.
• Contrast: clean black text on white paper works best; faint carbon copies and colored backgrounds are harder.
• Typed vs handwritten: printed text is recognized well; handwriting is generally not.

After recognition you see the average confidence. Below about 70%, the tool warns you that the scan is likely blurry, skewed or low resolution. Always proofread numbers, names and dates.

Columns, tables and headers are recognized as text but their layout is approximate in the TXT and Word outputs. The searchable PDF avoids that problem, because the visible page is the original scan.

Many PDFs mix typed pages with scanned attachments. With Skip pages that already have text turned on (the default for searchable PDF output), pages that already contain text are left alone and only image-only pages are recognized. That avoids doubled text and saves time.

For long files, type a page range such as 1-5 to process only what you need. Each page takes a few seconds, depending on your device.

The recognition engine runs inside your browser. Your scans are not uploaded to a server, which matters for tax forms, medical records, IDs and contracts. The first time you use a language, its recognition data (a few MB) is downloaded; after that it starts faster. If OCR fails on the first try, check your internet connection and try again.

If you are starting from paper, photograph the pages with Scan to PDF. It finds the page edges, straightens the perspective and cleans the background, and it can send the result straight here for OCR. If you only need an editable document from a PDF that already has text, PDF to Word is faster, and PDF to text extracts existing text instantly without OCR.