How to OCR a PDF (Free, No Upload)

Open PixelTools' OCR PDF tool and drop in your scanned PDF — an old contract, a scanned form, a receipt, or a photographed page, anything with no real text layer underneath. By default every page runs, but you can pick specific pages from the thumbnail grid or set a page range on longer documents.

Click Run OCR and recognition happens locally in your browser, page by page, with a live progress bar tracking real progress (not a frozen placeholder) — you can cancel at any point. When it finishes, download a searchable PDF (your original scan with an invisible, selectable text layer added) or the recognized text as a plain .txt file, whichever you need.

These solve different problems. A plain text extractor (like PixelTools' PDF Text Extractor) pulls text that already exists inside a PDF — it's instant, but only works if the file already has a real text layer, the kind produced by exporting from Word or a browser.

OCR PDF is for the opposite case: a scanned or photographed PDF that's really just pictures of text, with nothing to extract yet. This tool reads the pixels on each page and recognizes the words in them, which takes longer but works where a plain extractor can't. If you're not sure which one you need, try the faster text extractor first — it'll tell you plainly if there's no text layer to pull from.

A searchable PDF looks identical to your original scan, but with an invisible layer of recognized text placed over it — you can now select, search, and copy from it, and it's indexed properly if you search your files later. Pick this when you want to keep the document looking like the original (a signed contract, a form with a letterhead) but make it usable like text.

Plain text is the better choice when you just need the words — to quote a passage, feed it into another tool, or paste it into an editable document. It drops all of the original page's visual formatting and gives you the recognized words as a single .txt file instead.

OCR accuracy depends heavily on scan quality, since the tool is reading shapes on the page rather than understanding meaning. A clean, straight, high-resolution scan — roughly 300 DPI — typically comes out very accurate. A blurry phone photo, a skewed scan, or messy handwriting produces more mistakes, so always double-check anything important before relying on the result.

Recognition currently supports English text. The recognition model downloads once on first use in your browser session, so running OCR on a second PDF in the same session is faster than the first. Under the hood, each page is rendered to a canvas at roughly 144 DPI before recognition and processed by tesseract.js's LSTM neural network model rather than its older, less accurate legacy engine.

Making a filing cabinet's worth of old scanned contracts and forms searchable and copyable instead of stuck as flat images. Recognizing text in a phone photo of a book page, whiteboard, or printed document so you can copy it out instead of retyping it by hand. Archiving scanned receipts and records properly so they're indexed and searchable rather than just pictures sitting in a folder. Recovering readable, selectable text from scanned academic papers, articles, or handouts that predate digital text layers — anywhere a document exists only as a scan or photo, OCR is what turns it back into usable text.