PDF OCR

Extract text from scanned PDFs and image-only documents using optical character recognition (OCR), processed entirely client-side with no upload. Supports recognition in specific languages, making it useful for digitizing old contracts or photographed documents.

Drag and drop a PDF file, or click to choose one

Supports PDF files

Usage examples

Pulling text out of a scanned paper document

After scanning or photographing a paper document into a PDF, recognize the text inside it so you can copy and edit it.

Reading a PDF with no text layer

OCR can extract content from image-only PDFs that a plain PDF-to-text tool can't handle, since there's no embedded text to extract.

Recognizing a document in a specific language

Choose the matching language, like English, Korean, or Japanese, to improve recognition accuracy for that document.

Digitizing an old scanned contract

Run OCR on a scanned agreement saved years ago as an image-only PDF so you can finally search, copy, and quote its clauses.

Extracting text from a photographed whiteboard or sign

If you saved a photo of a whiteboard or printed sign as a PDF, recognize the text in it instead of retyping everything by hand.

Frequently asked questions

Is my uploaded PDF sent to a server?

No. Converting the PDF to images and recognizing text both happen entirely in your browser, and the uploaded PDF or its content is never sent to a server. The one exception is that the recognition model file for your chosen language is downloaded once from a public CDN on first use — this is just a generic program file, unrelated to your document's content.

Can I use this on a regular PDF that already has text?

You can, but if the PDF already has a text layer, our PDF-to-Text tool will be faster and more accurate. OCR is meant for image-only PDFs, like scanned documents.

How accurate is the recognition?

It depends on scan quality, font, and sharpness. High-resolution, crisp documents recognize more accurately, while handwriting or blurry scans may recognize less accurately.

How long does a long PDF take?

Pages are recognized one at a time, so the time scales with the page count. You can cancel at any point during processing if it's taking too long.

Does it recognize handwriting?

Not reliably. The recognition model is trained for printed text, so handwritten notes, cursive signatures, and stylized fonts often come out garbled or incomplete, even when the scan itself is sharp.

What if a page has text in more than one language?

You can only pick one recognition language per run, so a page mixing two languages will have the non-matching portions misread. If a document has distinct language sections, it's best to crop or split it and run OCR on each part with its matching language.

Do I need an internet connection to use it?

Only the first time. Once the language model has been downloaded and cached by your browser, you can reuse the same language for OCR without a network connection, as long as you don't clear your browser's site data.