OCR PDF online - recognise text in a scanned PDF

Try selecting text in your PDF viewer. If you can, the file has a text layer and OCR is unnecessary: PDF to text is faster and makes no recognition errors. If the whole page selects like an image, it is a scan and needs OCR. The tool checks up to 5 pages and warns you when the document already has text.

For a document containing both scans and digital text, choose only the scanned pages in the range field. Review names, dates and amounts against the original before using recognised text in another document.

Each selected page is rendered at 300 DPI (an A4 page is about 2480 × 3508 px) and recognised with the Tesseract engine (Apache-2.0) using LSTM models for the selected languages. Your file is not uploaded. Only the OCR engine (about 1.4 MB) and the language data (Polish 2.6 MB, English 2.9 MB, German 1.3 MB, Ukrainian 2.1 MB) are downloaded from the jsDelivr CDN. On later visits the browser usually loads them from its cache.

A searchable PDF looks like the original scan but has an invisible text layer under the image, placed where the words were recognised. Search (Ctrl+F), text selection and indexing then work, while the look of the pages stays the same. The option is not available for password-protected files, which can be read and recognised but not re-saved in the browser.

Processing time depends on your device, the number of pages and the number of selected languages. For long documents choose a page range. Best results come from straight, high-contrast 300 DPI scans; for faded print enable image enhancement.

For every page the tool shows the engine’s average confidence. Above 80% the text usually needs only minor corrections; below 60% expect many errors. The most common causes are skewed or blurry scans, very small print, coloured backgrounds and a missing language. Without Polish selected, letters such as ł or ę are read as t and e. Single photos and screenshots are easier to handle with the image to text tool.