OCR a Scanned PDF Online

PDFBell's OCR PDF tool reads the text sitting inside a scanned or photographed page — a paper document run through a scanner, a photo of a printed page, or any PDF built from images rather than a real text layer — and recognizes it as actual, selectable text. Pages like this look like a normal PDF but contain nothing a computer can search, copy, or index; OCR (optical character recognition) is what turns the picture of the words back into the words themselves.

Recognition runs entirely inside your browser tab using PaddleOCR, an open-source detection-and-recognition engine compiled to WebAssembly (ONNX Runtime Web) — nothing about the file, or the OCR engine itself, is fetched from or sent to a third-party server. Pick the language printed on the page — English, French, German, Spanish, Portuguese, Italian, Dutch, Chinese (Simplified or Traditional), or Japanese are all supported — and the matching recognition model downloads once and is cached for next time. Each page is rendered at a higher resolution than an on-screen preview, then processed one page at a time, with a progress bar tracking exactly where the run is.

Two outputs are available. The default, a searchable PDF, keeps every page looking exactly like the original scan but adds an invisible text layer underneath it, positioned line by line from the exact quadrilateral PaddleOCR detected for each line of text — including pages that carry their own rotation, so a page you turned 90° or 180° with rotate-pdf still gets an invisible text layer that lines up correctly once a PDF viewer re-applies that rotation. The page still looks like a photograph, but you can now select, search, and copy text from it in any PDF viewer. Turning that option off instead gives you the recognized words as plain text, useful when you just need to pull the content out rather than keep the scanned page's appearance.

OCR works best on clear, well-lit, printed text — a flatbed-scanned page or a clean phone photo of a document typically recognizes accurately. Low-contrast scans, heavy skew, handwriting, and unusual fonts reduce accuracy, the same limitation every OCR engine shares, not something specific to this tool. If a PDF already has a real text layer, pdf-to-text extracts it directly and is faster — this tool is specifically for the pages that don't.