Make a PDF searchable

Limits

• OCR works best on printed text in a sharp, straight, well-lit image. Handwriting, stylised fonts, low resolution and heavy shadows give poor results. Each result shows a confidence score; treat it as a guide, not a guarantee.
• The engine and English data (about 8 MB) are downloaded once from this site after you press the Download button. Hindi data (about 1.1 MB) is fetched the first time you choose Hindi. Your files are never uploaded. The data lives in your browser (Cache Storage and the normal HTTP cache), so whether it also works fully offline depends on your browser; clear site data to remove it.
• At most 50 pages are recognised per run (use the Pages box for the next batch). Pages are drawn at about 200 dpi, one at a time, to keep memory low.
• Every page is re-drawn as a JPEG at the chosen dpi, so the output no longer contains the original vector text, links, forms or bookmarks, and can be bigger or smaller than the input.
• The hidden text is placed from the recognised word boxes; selection highlights can be slightly off, and misread words are searchable as misread.
• Hindi text uses the bundled Noto Sans Devanagari font for the hidden layer; Latin text uses Noto Sans. Characters those fonts lack are left out of the hidden layer.

Ask the AI chat

The AI chat can run this tool for you. Try:

• “make this scanned pdf searchable”
• “add a text layer to my pdf”
• “OCR this pdf and keep it as a pdf”