How to copy text from an image or a scan

A screenshot of an address. A photographed page of a book. A scanned contract that came back from the other side as an image PDF. The text is right there in front of you, and yet you cannot select a single word of it, because as far as the computer is concerned it is a picture of text, not text.

The fix is OCR, optical character recognition, and it has quietly become very good. What used to produce "rn" for "m" and mangle anything with accents now reads ordinary documents almost perfectly, provided you feed it a reasonable image.

What "reasonable" means

OCR quality is mostly decided before the software ever runs, by the picture itself:

• Sharpness beats resolution. A crisp phone photo works better than a blurry high-resolution scan. If you can zoom in and read the letters comfortably, the OCR probably can too.
• Straight-on beats angled. Perspective distortion is the single most common reason for bad results from phone photos. Shoot from directly above the page.
• Even light beats bright light. A shadow of your hand across the text does more damage than a slightly dim room.
• Print beats handwriting. Printed text of any ordinary font reads well. Handwriting is a genuinely harder problem, and results vary from good to hopeless depending on the hand.

Extracting the text

1. Open Image to text and drop in the photo, screenshot or scanned PDF.
2. Run the recognition and give it a moment; a multi-page scan takes a few seconds per page.
3. Check the output next to the original, fix anything questionable, and copy or download the result.

The recognition runs on our server using a current-generation OCR model, which is what makes the accuracy possible; the file is processed and immediately deleted, and never stored. The tool is free with a daily allowance that covers normal use.

One detail worth knowing: the tool can rebuild paragraphs for you. A scanner sees lines, not paragraphs, so raw OCR output tends to break at the end of every printed line. The paragraph mode joins those fragments back into flowing text, which is usually what you want for anything longer than a receipt.

After the extraction

What you do next depends on the document. For a quick address or a phone number, copy and paste and you are done. For a document you need to edit properly, paste the text into your editor and reformat. If you are dealing with translations or academic work billed by length, Word counter tells you exactly what you extracted.

Frequently asked questions

How accurate is OCR on a normal document?

On a sharp, straight photo of printed text, expect nearly every word to come through correctly, with the odd slip on unusual names or numbers. Always skim the result against the original before using it anywhere that matters.

Does it read handwriting?

Neat, separated handwriting sometimes works; cursive mostly does not. OCR models are trained overwhelmingly on print, and handwriting recognition remains a much harder problem.

Is my document uploaded anywhere?

Recognition runs on our server, because the models that produce good results are too heavy for a browser. The file is processed, the text is returned, and the file is deleted immediately. Nothing is stored.

Can I extract text from a PDF scan with many pages?

Yes, drop the PDF in and each page is recognised in order. The whole document counts as one use of the daily allowance.

Why does my extracted text have a line break after every line?

Because the OCR sees the printed lines, which end where the paper ends, not where the sentence does. Switch the tool to paragraph mode and it joins the lines back into flowing text.