OCR a PDF
Read the text out of scanned pages, entirely on your device. Nothing is uploaded.
Choose files
or drop them here
Ctrl+V to paste from the clipboard
up to 1 file, 25.0 MB each
Stays on your device. Nothing is uploaded.
How to ocr pdf
- Choose one PDF, or drop it on the page.
- Pick the render resolution. 150 DPI reads well and keeps memory sane.
- Press Convert. Each page is drawn, then read on your device by OCR.
- Download the .txt file with the recognised text, page by page.
What OCR does here, and where it runs
Optical character recognition reads letters out of pixels. This tool renders each PDF page to an image, then hands it to Tesseract, the open-source OCR engine, compiled to WebAssembly and running entirely inside your tab. The language model (English) and the engine itself are served from this site: nothing about your document is ever sent anywhere.
It exists for the pages the text-extraction tool cannot help with: scans, faxes, photographed documents. The output is plain text with a marked section per page, ready for search, translation or an editor.
First use downloads the engine, a few megabytes, cached by the browser afterwards. Recognition speed depends on your device; a phone reads a page in a few seconds, a laptop faster.
Limits and honest notes
OCR accuracy follows image quality. Clean 300 DPI scans read nearly perfectly; phone photos of crumpled receipts will contain errors. 150 DPI is the sensible default; raise to 300 for small print.
English text only for now. Handwriting is not recognised: printed text only. Very long documents are best processed in page ranges.
One document per run, up to 25 MB. The recognition runs in this tab; the page images never leave your device.
Frequently asked questions
Why OCR instead of extracting text?
Extraction reads the text layer of a digital PDF. OCR reads the pixels of scans and photos. If extraction came back empty, OCR is the tool.
Which languages are supported?
English for now. More language models can be added later, each served from this site.
Why is the first conversion slow?
The OCR engine and language model download once (a few megabytes) and are cached. Subsequent runs start immediately.
Is my scanned document uploaded?
No. The pages are rendered and read inside your browser, on your device.
Private by design
Stays on your device. Nothing is uploaded. The conversion runs in your browser: open the Network tab while you convert and you will see that nothing is sent. No account, no watermark, no daily limit.