How to Extract Text from a Screenshot on Mac
Pasting a picture of an error message into a ticket makes it unsearchable, unquotable and useless to anyone reading it later. What you almost always wanted was the text.
Every capture is read the moment you take it, on your Mac, using Apple's Vision framework in its accurate mode. That means Copy Text is available immediately, with nothing to run and nothing to wait for.
Get the text out of a capture
-
Capture whatever contains the text
Press ⌃+⌥+A and drag over the error message, the table, the serial number, the code snippet.
Any capture mode works — the recognition is the same. -
Choose Copy Text
Rather than Copy, which puts the image on your clipboard, Copy Text puts the recognised words there instead. Paste into a ticket, a terminal, a chat, a search field — anywhere an image would be rejected.
The words, rather than a picture of the words. -
Or copy only part of it
The OCR view shows the recognised text blocks over the capture, so you can select the part you want. Copy All and Copy Selection are separate actions, which matters when the capture holds a whole screen and you need one line from it.
-
Run it on an image you did not capture
Snip Studio exposes App Intents to the Shortcuts app, including one that runs text recognition on whatever is on your clipboard. Copy a picture from a message, a document or a web page, run the shortcut, and you have the text.
Give that shortcut a keyboard shortcut in the Shortcuts app and it becomes a system-wide "turn this image into text" key.
An automation can do it on every capture, without asking.
Text pasted into a ticket is searchable by the next person who hits the same error, quotable in a reply, and translatable. A screenshot of that text is none of those things. It is worth the extra click almost every time.
What it recognises well, and what it does not
Recognition is very good on the thing screenshots are usually made of: interface text, at normal contrast, in a normal typeface. It gets steadily less reliable on:
- Very small type. Capture at a larger zoom level rather than capturing the small version and hoping.
- Low contrast. Grey-on-grey placeholder text is genuinely hard.
- Unusual or decorative typefaces, and heavily stylised logos.
- Text at an angle, as in a photograph of a screen rather than a screenshot of one.
Recognition is case- and diacritic-aware, so accented characters come through correctly rather than being flattened.
Getting text out of something longer than the screen
A table that runs past the bottom of the window is the case where this stops being a convenience and starts being the only reasonable option. Retyping forty rows is not a plan, and screenshotting them in three overlapping pieces gives you three images to reconcile.
Capture the whole thing as a scrolling capture instead. Snip Studio stitches the page into one image, and the text recognition runs over the stitched result — so one Copy Text gives you every row, in order, from a page you never saw all of at once. Scroll capture is part of the paid ULTRA upgrade.
The same applies to a long terminal buffer, a chat thread, or a settings pane with more rows than height. Capture it whole, copy the words, and let the layout be somebody else's problem.
Getting a usable result out of a bad source
When recognition struggles, the fix is almost always upstream of the capture rather than after it. Three things work, in descending order of how much they help:
- Zoom in before capturing. Recognition works on the pixels it is given. Pressing ⌘++ in a browser or bumping a document's zoom before you capture does more than any amount of processing afterwards.
- Switch the source to a light appearance. Light text on a dark background is read perfectly well, but low-contrast dark themes — grey on near-black — are the common failure case.
- Capture a tighter region. A capture holding one paragraph gives you a clean block. A capture of a whole display gives you the surrounding interface chrome mixed in with what you wanted.
The same recognition does two other jobs
It makes your library searchable. Because the text was read at capture time, you can find a screenshot later by a phrase that appeared inside it — see how to search screenshots by their contents.
It finds what is sensitive. Email addresses, URLs and file paths are pulled from the same recognised text and offered for redaction, alongside faces and codes found in the pixels. That is covered in how to blur or redact sensitive information.
One pass, three uses, all of it on your machine.
Common problems
The recognised text has mistakes
Usually small or low-contrast source text. Capture at a higher zoom level — recognition works on the pixels it is given, and more pixels is the single biggest improvement available.
Copy Text is empty
There was no recognisable text in the capture — a photograph, an icon, a chart with no labels. Nothing is wrong.
Line breaks come through oddly
Recognition preserves the visual layout, so a two-column table becomes two blocks rather than interleaved rows. Copy the columns separately if the order matters.
You wanted the image and got text
Copy and Copy Text are separate actions. Both are available on the same capture; neither replaces the other.
The Shortcuts action is not listed
Open Snip Studio at least once after installing so its App Intents register, then reopen the Shortcuts app.
Is any of it sent anywhere?
No. Recognition runs locally through Apple's Vision framework. See the Privacy Policy.
Frequently asked questions
How do you copy text from a screenshot on a Mac?
Take the capture in Snip Studio and choose Copy Text. The recognised words go to your clipboard instead of the image, ready to paste anywhere.
Can you extract text from an image you already have?
Yes. Snip Studio provides a Shortcuts action that runs text recognition on whatever is on your clipboard, so it works on any image, not only its own captures.
Does the text recognition work offline?
Yes. It runs on your Mac through Apple's Vision framework, so it works with no network connection and nothing is uploaded.
Why is some text recognised incorrectly?
Very small, low-contrast or decorative text is harder to read. Capturing at a larger zoom level usually fixes it.
Can you copy just one part of the recognised text?
Yes — the OCR view shows the recognised blocks over the capture, and Copy Selection copies only what you highlight.
Recognition behaviour, the OCR view and the Shortcuts integration are from Snip Studio's own documentation. Local processing matches the published Privacy Policy. Last checked 1 September 2026.