⇧⌘2 draws a box over anything on screen and copies the words inside it.
Text that cannot be selected — a video frame, a scanned PDF, a screenshot someone sent you — is still text. Capture reads it off the screen.
Recognition happens on your Mac. Nothing is uploaded, and it works offline.
The first capture asks for Screen Recording and shows the same drag-to-grant window as the install. macOS requires it for anything that reads the screen.
An image already in your history carries its recognised text too. Open its preview and press ⇧⏎ to paste the words instead of the picture, or use Copy Text.
Esc cancels the crosshair, and nothing is copied.
Did this answer it? If not, write to [email protected] and I will fix the page.