Make a scanned PDF searchable

SwiftPDFEdit's OCR tool reads the printed text inside a scanned PDF and adds an invisible text layer over the page, turning a picture of a document into one you can search, select and copy — and it does the recognition inside your browser rather than on a server. The scan looks exactly as it did before; what changes is that the words underneath are now real text.

A scanned page is a photograph. Your PDF reader sees pixels, not characters, which is why Ctrl+F finds nothing, why you cannot copy a paragraph out of it, and why no editor can change a single word. Optical character recognition is the step that fixes that: it looks at the shapes on the page, works out which characters they are, and records both the text and exactly where on the page each word sits.

Every other tool that does this uploads your document. That is a real problem when the scan is a contract, a medical letter, a bank statement or a passport page — the files people most often need to make searchable are precisely the ones they should be most careful about sending to a stranger's server. SwiftPDFEdit runs the Tesseract OCR engine compiled to WebAssembly inside your own browser tab. The language models are served from our own site, cached by your browser, and the page images are never transmitted. You can disconnect from the internet after the first run and it still works.

The output is a normal PDF that opens anywhere. Underneath each recognised word we write the same word in an invisible text layer, positioned over the ink it came from, so the file still prints and looks identical while search, text selection and copy all behave the way they do in a born-digital document. Because the text is genuinely in the file, SwiftPDFEdit's own editor can then work on it too — you can find and replace across the scan, redact a name out of it, or edit a line of the recognised text in place.

Be realistic about what OCR can and cannot do, because we would rather you knew up front. It reads print well and handwriting badly — no engine, ours or anyone else's, reliably reads handwriting. A crisp 300 DPI scan gives excellent results; a dim phone photo taken at an angle gives poor ones, and we show you the confidence score for each run so you find out from us rather than from a search that comes back empty. Recognition preserves words rather than layout, so columns, tables and forms become searchable but their reading order may not match the visual order. It takes a few seconds per page — that is real work happening on your machine, not an upload queue.

OCR is part of the Pro plan ($10/month or $49/year), which also includes editing a document's original text, unlimited documents and no daily task cap. The core tools — merging, splitting, compressing, converting, rotating and the annotation editor — remain available on the Free plan.