Second ReadHow do you copy text from an image?
15 September 2026
Text inside an image is pixels, not characters, so no amount of dragging your cursor will select it. Getting it out means having something read the picture. Here is what actually works, ordered by how little effort each takes.
You copy text from an image by running optical character recognition, usually called OCR, over the picture. OCR looks at the shapes in the image and works out which letters they are, producing text you can select, paste, and search.
Every major platform now ships this. On a Mac, open the image in Preview or Quick Look and drag across the words; macOS Live Text makes them selectable. On iPhone and iPad, open Photos, tap the small text icon in the corner, and select. On Windows, the Snipping Tool has a Text Actions button that copies all text in a capture. In Chrome on Android, Google Lens is built into the long-press menu on any image.
What all of those have in common is that they work on an image you already have. If the text is on your screen right now and you want it kept rather than pasted, a browser extension that captures and stores it in one action saves the round trip through a screenshot file. Second Read is one, free, and reads the text on your own machine.
Why can I not select text on a website?
Text on a website is usually unselectable for one of three reasons: it is an image rather than text, the page has disabled selection in CSS or JavaScript, or it is rendered inside a canvas element such as a web reader or a slide viewer.
The CSS case is the easiest to spot, because the text looks perfectly crisp when you zoom in and the cursor still refuses to catch. That is a site setting user-select to none, and it is trivially defeated by turning off JavaScript or using Reader mode.
The other two cases produce genuinely unselectable pixels. Kindle Cloud Reader, most PDF viewers embedded in a page, Google Slides in presentation mode, Instagram and Twitter screenshots, and any book scan fall here. Those need OCR, not a workaround.
Is there a way to copy text from an image without uploading it anywhere?
Yes. Modern OCR runs entirely on your own device, with no server involved. The macOS, iOS, and Windows tools above all do the recognition locally, and so does Second Read, which uses Tesseract compiled to WebAssembly and runs it inside your browser.
This matters more than it sounds. Most free "image to text" websites work by uploading your picture to their server. If the image is a page of a book, a contract, a medical letter, or a screenshot of a private conversation, you have just handed a copy to a company whose retention policy you have not read.
The way to check any tool is to turn off your wifi and try it. If it still works, the recognition is happening on your machine.
How accurate is OCR on a screenshot?
OCR on a screenshot is close to perfect, and considerably better than OCR on a photograph. A screenshot has no perspective distortion, no blur, no uneven lighting, and pixel-aligned type, which are the four things that break recognition.
Where it still struggles: text under about ten pixels tall, light grey text on a white background, heavily stylised or script typefaces, text over a busy photograph, and columns that the reading order gets wrong. Zooming the page in before you capture fixes most of these, because it hands the recogniser more pixels per letter.
For ordinary body text at ordinary zoom, expect to fix a word every few paragraphs at most, which is why any tool worth using shows you the result in an editable box rather than saving it silently.
How do you copy text from a PDF that will not let you select it?
A PDF that will not let you select text is almost always a scan: a photograph of pages wrapped in a PDF container, with no character data inside it at all. The fix is to run OCR over it, which most PDF readers can do for you.
Acrobat has Scan and OCR, Preview on macOS can select text in scans directly through Live Text, and on Windows the Snipping Tool works on whatever is on screen regardless of the file type. If you only need a paragraph rather than the whole document, capturing the region on screen is faster than converting a two hundred page file.
The other possibility is that the PDF has a copy restriction set. That is a flag rather than encryption and it only asks the viewer politely to refuse; screen capture with OCR is unaffected by it.