A scan or a photo of a document is an image: the text you can see cannot be selected, searched or copied. OCR recognises it and hands it back as real text.
You can ask for a plain text file or, better, a searchable PDF: the same picture as always, with an invisible text layer underneath. It looks exactly the same, but now you can search it and copy from it.
The recognition is done by Tesseract compiled to WebAssembly, running inside the tab. A scanned contract or invoice is not sent to a service to be read, which with that kind of document is not a small detail.
How to use it
- Drop in the scan, the photo or the PDF.
- Choose the document language: getting it right changes a lot.
- Decide between plain text and a searchable PDF, then process.
Frequently asked questions
- Why do I have to state the language?
- Because recognition does not work letter by letter: it uses a language model to choose between possible readings. With the wrong language, “word” can come back as “vvord”.
- Can I tick two at once?
- Yes, for bilingual documents. There is a cost: each added language makes the job slower and adds some confusion.
- The result has mistakes — what can I do?
- OCR depends mostly on the quality of the original. Raise the resolution to 300 dpi, straighten the image if it is skewed, and check the contrast is good.
- Does it read handwriting?
- Essentially no. Tesseract is trained on printed text; with handwriting the results are not usable.
Related tools
Compress PDFRe-encodes the images inside without touching the text or structure.
Organise pagesSee every page, drag them about, turn them and drop the spares.
PDF to imagesOne image per page, at whatever resolution you need.
PDF to textPulls out the document’s text layer, without OCR.
Extract images from a PDFPulls the embedded photos and graphics out as separate files.