This tool reads the text inside a scanned PDF, a document that's really just a picture of text, with nothing selectable or searchable, and produces a new PDF with an invisible, searchable text layer placed exactly over the original image, plus a plain .txt file of everything it recognized.
It uses Tesseract, a genuinely capable open-source OCR engine, running entirely in your browser via WebAssembly. The first time you use a given language, it downloads a roughly 15MB language model, which is then cached so every OCR run after that is instant to start, we're upfront about that download rather than hiding it behind a silent delay. English and Hindi are supported alongside several other languages, and you can OCR just the pages you select instead of an entire long document.
Recognition quality depends heavily on scan quality: crisp, high-contrast scans of typed text OCR very accurately, while blurry photos, tight handwriting, or low-contrast faxes will have more errors, this is a real limitation of OCR technology generally, not something specific to this tool. Because it runs client-side in a Web Worker, your document is never uploaded, which matters for anything scanned that shouldn't leave your device, IDs, medical paperwork, signed contracts.
The most common reason people need this is a paper document that only exists as a phone photo or a flatbed scan, an old contract, a printed article, a form filled out by hand and then scanned, where you need to search for a specific word, copy a paragraph, or just make the file accessible to a screen reader. Once OCR'd, the searchable PDF behaves like any normal digital document for those purposes, while the plain-text export is useful for pasting content straight into an email or another document without retyping it.
How to OCR Scanned PDF
- Upload your scanned PDF.
- Choose the document's language from the picker.
- Select which pages to OCR (or leave all pages selected).
- Run OCR and download the searchable PDF and/or the plain text file.
Frequently asked questions
Which languages are supported?
English and Hindi at minimum, plus several other major languages: pick the one that matches your document before running OCR.
Why does it download something the first time I use it?
OCR needs a language-specific recognition model (about 15MB) the first time you use that language. It's cached afterward, so every later use starts instantly.
Is my scanned document uploaded to recognize the text?
No. Recognition runs entirely in your browser in a background Web Worker, so your file is never sent to a server.
How accurate is the OCR on handwriting?
OCR generally works best on typed or printed text with a clean scan. Handwriting recognition is much less reliable: expect more errors there than with printed text.
Can I OCR only some pages instead of the whole document?
Yes, use the page picker to select just the pages you need recognized, which also speeds up longer documents.
Related tools
- Compress PDF: Shrink file size with five presets, or target an exact size.
- Add Text & Images: Place text boxes and images anywhere on a PDF page.
- Annotate & Highlight: Highlight, draw and add sticky notes on any page.
- Summarize PDF: Generate a concise AI summary of a long document.