What is OCR and how to use it on a PDF
OCR, short for optical character recognition, reads the letters on a scanned page and turns them into real text you can search and copy. ShiftAny Files runs it in your browser and gives you a text file or a searchable PDF.
Published
What OCR actually does
A scan is a photograph of each page, so the PDF does not know which letters appear in it. OCR looks at the picture, recognises the characters and writes them out as text. The result can sit in a plain text file or as an invisible layer over the original image.
When you need it
You need it when you cannot select any text in a PDF, when the search shortcut finds nothing although you can see the word on screen, or when all you have is photographs of paper documents.
Step by step
- Add the scanned PDF, or photos of the document in JPG, PNG or WebP.
- Choose the language the text is written in, and whether you want a text file or a PDF with selectable text.
- Press the recognise button and wait: each page takes a few seconds.
- Download the result and read it through before you use it.
How it works here
Each page is drawn at 200 dots per inch and recognised with Tesseract, an engine that runs inside your browser through WebAssembly. Its language data is served from this same site, so the document is never sent anywhere. Spanish and English are supported. In a searchable PDF the original image stays visible and the recognised text sits behind it.
Getting better results
Straight, well-lit scans with sharp print give the most accurate text. Pick the language that matches the document, and double-check names, figures and dates, because recognition is never perfect.
Recognise the text of your own scan with the tool this guide is based on:
OCR PDF
Recognize the text in a scanned PDF or in a photo of a document.
Frequently asked questions
Does OCR read handwriting?
It is built for printed text. Handwriting usually comes out unreliable, so do not expect a usable result from it.
Which languages can I recognise?
Spanish and English. Choosing the language that matches the document matters, because the engine uses it to decide which characters are likely.
Is the recognised text always correct?
No. Blurry scans, unusual fonts and small print produce mistakes, which is why every result should be checked before it is used.