OCR PDF reads the words on each page of a scanned PDF and adds them back as an invisible text layer over the original image. The picture of the page is not changed at all. Afterward you can search the document in any PDF reader, select and copy the text, and find the file by its contents on your computer. The recognized text is also shown on the page so you can copy it straight away.
The recognition is done by Tesseract, an open-source OCR engine compiled to WebAssembly so it runs inside your browser. The first run downloads about 5 MB of engine and English model, which your browser then keeps cached. Your PDF is never uploaded: each page is rendered at up to 200 dpi on your own device, read there, and the words are written into a new copy of the file. Close the tab and nothing of your document remains anywhere.
Only English is recognized at the moment. Accuracy depends on the scan: clean, straight pages of printed text come out well, while faint copies, skewed pages, unusual fonts and handwriting produce gaps. Words the engine is unsure about are left out of the hidden layer rather than guessed, so a search can occasionally miss a word that is plainly visible on the page. One PDF at a time, up to 100 MB.
How to use OCR PDF
- 1Select the scanned PDF, or drop it onto the page. The tool works on one file at a time.
- 2Leave Language of the text set to English (in your browser), the only language available right now.
- 3Keep Skip pages that already have text checked unless you want every page re-read; pages that already have a text layer are otherwise left exactly as they are.
- 4Click Make searchable and wait while each page is rendered and read. The progress bar names the page being worked on.
- 5Check the Recognized text preview, use Copy if you only need the words, then click Download to save the new file, named after the original with -searchable added.
Frequently asked questions
- Is my scanned PDF uploaded to a server?
- No. English recognition runs on your own device with Tesseract compiled to WebAssembly. The only thing downloaded is the recognition engine and language model, about 5 MB from freeaipdf.com, which your browser caches for next time. Your file never leaves the browser.
- Does OCR change how the pages look?
- No. The page images are kept exactly as they are. An invisible text layer is placed over each recognized word, sized and positioned so that a selection highlight lands on the printed word. The result is saved as a new file, so the original on your computer is not touched.
- Can it read handwriting?
- Not reliably. The English model is trained on printed text, so handwritten pages mostly come back with gaps or wrong words. Words the engine is not confident about are left out of the text layer rather than guessed, which keeps searches from matching garbage but means handwriting is largely skipped.
- Why does it say every page already has a text layer?
- Skip pages that already have text is checked by default, and every page in your file already contains selectable text, for example a PDF exported from Word. That file is already searchable. If you want the pages read from the image anyway, uncheck the box and click Make searchable again.
- Which languages does it recognize?
- English only at the moment, because that is the only model shipped to the browser. Text in other languages that use the Latin alphabet is read with the English model and will contain more mistakes. Words in other scripts are skipped in the hidden layer even where the engine picks them up, because that layer uses a built-in font that covers Latin characters only; the page image still shows them.
- How long does it take?
- It depends on your device and the number of pages. Each page is rendered and read in turn on your own processor, so a long scan on an older laptop or a phone takes a while, and the engine uses roughly 100 MB of memory until the run finishes. Pages skipped for already having text cost almost nothing.