FREE · PDF TOOLS
Searchable PDF OCR.
Add local OCR text layers to scanned PDF pages for search and selection. Review results and retain all original pages, including Arabic and mixed RTL text.
20 MiB source · 200 source pages · up to 20 selected pages per run. All original pages stay in the exported PDF. Documents stay in your browser; OCR engine and selected models load from this site. Existing digital signatures become invalid. No PDF/A certification or secure redaction.
Printed text works best; handwriting is unreliable. Arabic and mixed Arabic/Latin scans use a local Unicode text layer; review word order, numbers and punctuation before export. Supplementary Unicode characters remain unsupported. Pages with any existing text are skipped, including pages mixing digital text and scanned areas.
Steps to use Searchable PDF OCR
Choose a PDF and up to three supported recognition languages. Recognize up to 20 selected pages, review the text and exclude incorrect layers, then prepare and download the searchable PDF.
How Searchable PDF OCR processes your input
Eleven recognition models are available: English, French, Arabic, Spanish, German, Portuguese, Italian, Dutch, Hindi, simplified Chinese and Japanese. Native invisible layers are used for LTR text; Arabic and mixed RTL/Latin words use a local Unicode bidi layer positioned from OCR word boxes. Both paths align to original visible page geometry, including CropBox and rotation. All source pages remain in the output; excluded and unselected pages receive no layer. Any existing nonempty text skips the whole page, including mixed digital/scanned pages. Text layers are read-only: rerun OCR or exclude incorrect pages; use PDF-to-Word for manually corrected text. Extra rotation and contrast affect recognition only; no automatic deskew. Generated fixture checks verify logical text, mixed numbers and punctuation, approximate word selection bounds and unchanged pixels with PDF.js; general accuracy and native viewer behavior need review. Existing digital signatures become invalid. All processing is local.
Searchable PDF OCR: worked example
Recognize a scanned reference number, keep a digital page unchanged and find the reference in the exported PDF.
Make a scanned PDF searchable
- Choose a PDF (up to 20 MiB and 200 pages) and up to three recognition languages.
- Select up to 20 pages to recognize in one run.
- Review the recognized text and exclude any page whose result is wrong.
- Prepare and download the searchable PDF: the pages look the same, but you can now search and select their text.
Example: recognize a scanned reference number, keep a digital page unchanged and find the reference with the search box of your PDF viewer.
How the text layer works
- An invisible text layer is placed over each recognized page, aligned with the visible page, including cropped and rotated pages. All original pages stay in the file.
- Eleven languages are available: English, French, Arabic, Spanish, German, Portuguese, Italian, Dutch, Hindi, simplified Chinese and Japanese. Arabic and mixed right-to-left text use a word-positioned layer.
- Pages that already contain text are skipped. Printed text only: handwriting is unreliable. The layer cannot be edited; for corrected text, use PDF to Word.
- Existing digital signatures become invalid, and the output is not PDF/A or a redaction. Everything runs in your browser. For photos and screenshots, use image to text.
Questions about Searchable PDF OCR
Will my PDF look different after OCR?
No. The page images are unchanged; only an invisible text layer is added for search and selection.
Can it read handwriting?
Not reliably. The models are made for printed text.
Is my PDF uploaded?
No. Recognition runs in your browser with models served by PasteZap.
Is Searchable PDF OCR free, and do I need an account?
Yes. This tool is free to use in your browser without an account. Processing takes place on your device.
What are the limits and what should I check?
20 MiB source, 200 source pages, 20 selected pages per run and 500,000 review characters. Arabic/RTL is supported with a local word-positioned Unicode layer; supplementary Unicode remains unsupported. Printed text only; handwriting is unreliable. Pages with existing text are skipped. No PDF/A, signature preservation or secure redaction. Review the downloaded document, especially page order and watermarks.
Is my input sent to a server?
Your input is processed in your browser and is not uploaded by this tool. Files and pasted text are cleared when the page is refreshed. Normal website requests are still required to load the page and its libraries. Read the privacy explanation.