Turn a scanned PDF or photo into text
Updated October 29, 2026
A scanned PDF looks like a document but behaves like a photo: you cannot search it, select a sentence or copy an address. Optical character recognition (OCR) reads the letters in the picture and gives you the text back.
The OCR tool on this site runs Tesseract in your browser, so the scan stays on your device. Here is how to use it well.
From a scan or a photo to text
Drop the PDF or the image, choose the language of the text, and extract. For a PDF, every page is rendered and read in turn; the text appears with a page marker between pages. Copy it, or save it as a .txt file.
Pick the right language: the model knows the words and letter shapes of that language, and the accuracy drops a lot with the wrong one. For a document that mixes Romanian and English, tick "also English".
A searchable PDF instead of raw text
Tick "searchable PDF" and the tool adds an invisible text layer over each scanned page. The file looks exactly like the scan, but you can search it, select text and copy from it, and the words are indexed by desktop search. This is the format archives and law firms use.
Getting a clean result
OCR is only as good as the picture. A flat page, even light, no shadow from your hand and the text horizontal: that is 95 percent of the quality. Phone photos taken at an angle, in a dim room or with the page curved give wrong letters and merged words.
If a page is sideways, rotate it first with the organize tool. For a photo, crop to the text block before OCR; borders and background confuse the layout analysis.
Tips
- Scan at 300 dpi for small print; 200 dpi is enough for a normal letter.
- Handwriting is not supported; only printed text.
- The language models are 1.5 to 3 MB each and are fetched from this site once.
Questions
- Which languages are supported?
- English, Romanian, Spanish and Portuguese, alone or combined with English.
- Does the text keep its layout?
- Line breaks and paragraphs yes; columns and tables come out as reading order, not as a table.
- Is the scan uploaded?
- No. The OCR engine runs in a Web Worker in your browser.