How to use OCR PDF
- Click “Select PDF file” or drop a scanned PDF onto the box.
- Choose the language the document is written in. Recognition is far more accurate with the right language.
- Leave “Skip pages that already contain text” ticked so that only the scanned pages are processed.
- Press “Recognise text” and keep the tab open while the pages are read.
- Download the searchable PDF. The recognised text is also shown, ready to copy or to save as a .txt file.
About this tool
Each page without text is drawn at about 250 dots per inch and handed to Tesseract, an open-source recognition engine that also works inside many desktop scanning programs and runs here as WebAssembly in your browser. It finds the lines and words on the page and returns every word with its position. Those words are written into the PDF in text render mode 3 – invisible text – right on top of the printed words. The page images themselves are not re-compressed or altered, so the document looks identical and the file grows by only a few kilobytes per page.
The invisible layer is there for searching, selecting and copying: Ctrl+F in a PDF reader, your computer’s file search, or pasting a paragraph into an e-mail. It uses a font without letter shapes that only records which character sits where. That is why it works for every supported script, including Arabic, Urdu, Hindi, Chinese and Japanese, without embedding a large font. Pages that already have real text are left untouched, while a scan that merely carries a typed page number or a “scanned by” stamp is still recognised.
Recognition is good, not perfect. Clean office scans at 200–300 DPI in an ordinary typeface usually come out with few mistakes. Expect more errors with low-resolution phone photos, tilted or curved pages, faint photocopies, coloured or patterned backgrounds, very small print and decorative fonts. Handwriting is not recognised. Tables and multi-column pages are read, but the order of the text may not follow the columns. Pages must be upright – turn sideways pages with Rotate PDF first. Accuracy also differs by script: Latin and Cyrillic print is read best, clear Chinese, Japanese and Hindi print well, and Arabic and Urdu noticeably less reliably.
Speed depends on your device, because your processor does the work a server would do elsewhere. A recent laptop needs roughly one to three seconds per page and works on several pages at once; a phone is several times slower and should stay awake with the tab in front. The first run downloads the engine (about 1.5 MB) and the chosen language (0.7–3 MB) from this site; later runs use the copies in your browser’s cache. There is no page limit, and for documents of more than 50 pages you see a time estimate before you start.
Frequently asked questions
What does OCR do to my PDF?
It adds an invisible layer of text on top of the scanned pages. The pages look the same, but you can now search the document and select and copy its text, and screen readers and search tools can read it.
Is there a page or file limit?
There is no fixed limit and no daily quota. Recognition runs on your device, so long documents simply take longer: allow one to three seconds per page on a laptop and noticeably more on a phone.
Which languages can it read?
English, Spanish, French, German, Portuguese, Italian, Dutch, Russian, Arabic, Urdu, Hindi, Simplified Chinese and Japanese. Choose the language of the text before you start. One language is used per run, so a page that mixes two scripts is only read well in one of them. Latin and Cyrillic print is read most reliably; Arabic and Urdu are the weakest, especially in calligraphic Nastaliq typefaces.
Can it read handwriting?
No. The engine is trained on printed type. Neat block capitals are sometimes recognised, but joined-up handwriting comes out as nonsense.
Why are some words wrong?
The engine works out each word from its shape. Blurry, tilted, low-resolution or faint pages give it less to go on. Scanning again at 300 DPI, with the page flat and straight, is the most effective fix.
Why is the highlight a little off when I select text?
The invisible words are positioned by the box around each printed word, so a selection can be slightly taller or wider than the letters. Searching and copying are not affected.
Should I compress the PDF before or after OCR?
Either order works with Compress PDF in its default “Keep text and links” mode, which shrinks the page pictures and leaves the text layer in place. Only the “pages become pictures” mode would throw the recognised text away again, so avoid that one after OCR.
Are my PDF files uploaded to a server?
No. The whole process runs inside your browser with JavaScript. Your PDF files never leave your device, which also means the tool keeps working if you go offline after the page has loaded.
Is this tool really free?
Yes. There is no sign-up, no daily limit, no watermark and no paid tier. The site is supported by advertising shown on the page, not by charging for the tools.