How to copy text out of an image on a Mac

An error message you cannot select. A code snippet inside a video. A phone number in a photo. macOS can read text out of images — here is where that ability lives and where it does not reach.

In English3 min read

What macOS already does

Live Text has been part of macOS since Monterey. Open a picture in Preview, Photos, Quick Look or Safari and you can usually drag across text in it exactly as you would in a document. It handles photographs, screenshots and scanned pages, and it works offline — the recognition happens on your Mac.

If a drag starts a rectangular selection instead of a text selection, look for the small Live Text indicator in a corner of the image and click it to switch modes.

Where it does not reach

Live Text needs an image. Very often the text you want is not in an image at all — it is on screen, in a running app, and simply refuses to be selected:

  • Error dialogs, installers and system alerts that render text as an unselectable label.
  • Video and screen shares, where the frame is not a file you can open.
  • Remote desktop and VM windows — the text is a picture of text as far as your Mac is concerned.
  • Apps that deliberately block copying.

The workaround is always the same shape: capture the region as an image, then recognise it. The friction is in how many steps that takes — screenshot, find the file, open it, wait for Live Text, select, copy.

The Shotcake toolbox showing the on-screen colour picker magnifier and HEX value
Shotcake's toolbox sits alongside capture: text recognition, colour picker, ruler and pin, all one shortcut away.

Collapsing it into one step

Shotcake folds those six steps into two: select the region, and the recognised text is on your clipboard. Recognition runs through Apple's Vision framework on your Mac — the image is not uploaded to a service, which matters when the unselectable text is a licence key or a customer record.

On-device recognition is also why it works with no network connection at all, and why it costs nothing per use. There is no per-image quota because there is no server.

Getting better results

  1. Capture at native resolution. Do not scale a window down before capturing; recognition accuracy falls off quickly with pixel density.
  2. Crop tight. A region containing only the text you want beats a full-screen capture the recogniser has to scan.
  3. Watch out for low contrast. Grey-on-grey UI text is the most common failure. If the app has a light mode, switching to it usually fixes recognition.
  4. Check numbers and keys by eye. Any recogniser can confuse 0/O and 1/l. For anything you are going to paste into a terminal, read it back.

Keep reading

Take a better screenshot in the next ten seconds

One shortcut, one drag, and the annotation editor is already open.

Coming soon to the Mac App Store