OCR PDF
Recognises the words on each page and writes them back as an invisible layer sitting exactly over the image. The page looks identical and Ctrl+F starts working.
Local processingAccepts PDFUp to 100 MB
Drop your file here
PDF · up to 100 MB each · you can also paste from the clipboard
Nothing is processed until you press the button below.
How this tool works
- 1Each page is rendered to an image at a resolution high enough for recognition.
- 2Tesseract, compiled to WebAssembly, reads the words and reports where each one sits.
- 3A new document is built from the page images with those words drawn on top at zero opacity: invisible, but selectable and searchable.
What it will not do
- Accuracy depends almost entirely on the scan. Below roughly 200 dpi, or on skewed, shadowed or creased pages, results degrade quickly. Straighten the page first if it is crooked.
- English only in this build. Other languages need their own model.
- Handwriting is a different and much harder problem, and is not attempted.
- The output pages are images, so the file is larger than the original and the text layer is a best guess rather than a certainty. The average confidence is reported so you can judge it.
Where this runs
OCR PDF runs entirely inside this page. Your file is read by the browser, processed in memory and handed back to your downloads folder. It is never sent to a server, because there is no server involved in this tool.
Read the full processing model →Questions
No. The recognition engine and its language model are downloaded to your browser and run there. The document never leaves your device.