Route one: Live Text, no app needed
Open the file in Preview, Quick Look or Safari. Move the pointer over the text and the cursor turns into a text cursor. Select, copy, done. It also works on photos in Photos and on frames of a paused video in QuickTime.
Live Text needs the picture open in an app so you can select inside it. That is the limit worth knowing, because a screenshot you took with ⇧⌘4 goes straight to the clipboard and never becomes a file you can open.
Route two: copy first, recognise after
Recognition can also run on whatever is already in the clipboard. Raycast has text recognition, ClipBear does too, and dedicated capture apps read a region off the live screen. On-device recognition is not a rare capability in 2026, and any page telling you otherwise is selling something.
Recognition here is not cached when you copy. The words are read at the moment you paste, which is why a screenshot from last week reads fresh and nothing ever expires out of a cache.
The same recognition feeds search. Type a word that appears inside a screenshot and the screenshot comes back, which beats scrolling a column of thumbnails that all look the same size and shape.
Route three: one keystroke on the way out
In Klipto the recognised text is a modifier away from the ordinary paste:
1. Take the shot with ⇧⌘4, or copy any image out of a browser, Slack or Finder.
2. Open the popup with ⌥⌘V and select the image.
3. Press ⇧Enter.
The words land in whatever app has focus. Plain Enter or ⌘V still pastes the picture, so both live on one item and you never have to decide at screenshot time which one you will need later. Pulling the text out consumes nothing, so you can do it again next month.
Recognition runs on-device through Apple Vision. The image does not leave the Mac and no connection is required, which matters when the screenshot is of an invoice or an internal dashboard.
The trap: accuracy is Apple's, not the app's
Sharp black-on-white text comes back clean. Low-resolution screenshots, text rotated off horizontal, low contrast, heavy stylisation and thin condensed fonts all come back worse, and sometimes wrong in ways that read as plausible. Numbers are where that hurts, so check any figure you paste from a receipt.
The fix is usually the screenshot, not the software. Zoom in before you capture, straighten a photo of a page, take the shot at full resolution rather than scaling a window down first. Nothing on-device reconstructs detail the pixels never contained.
What I use
Live Text for anything already on screen. It is free and it is good. ⇧Enter for the case that actually fills my day: a screenshot I took ten minutes ago, still in history, whose text I now need in a commit message or a reply.
Klipto is where I built it because the image was already in the clipboard, so reading the text out of it removed a step instead of adding an app.