How to get the text out of a screenshot

All three systems now read text in images natively. On a Mac you can select text inside a screenshot as if it were a document.

Updated 14 September 2026

Mac: Live Text needs no tool at all#

Open a screenshot in Preview, Photos, or even Quick Look, and drag across the text with your cursor. It selects like text in a document. Cmd+C copies it.

This works everywhere images appear on a modern Mac — Safari, Messages, Notes, Mail. Point at a screenshot of an error message and copy the error code straight out of the picture.

It recognises many languages and handles mixed layouts, though multi-column text comes out in reading order rather than column order, which occasionally interleaves things oddly.

Right-clicking selected text also gives you Look Up, Translate, and — for a phone number or address — the option to act on it directly.

Windows 11: Snipping Tool text actions#

1. Capture or open an image in the Snipping Tool.
2. Click the text actions button in the toolbar.
3. Recognised text becomes selectable. Drag over what you want, or click Copy all text.

The same panel has Quick redact, which finds email addresses and phone numbers and blacks them out in one click. Useful before posting a screenshot anywhere public.

Windows 10 doesn't have this. Alternatives: PowerToys includes a Text Extractor (Win+Shift+T) that reads text from any region of the screen directly, no capture needed. ShareX also has OCR built in. Both are free.

Chromebook#

Google Lens is built into Chrome. Right-click an image, choose Search with Google Lens, then the text mode. Select and copy.

Google Drive does full OCR on uploads: upload the screenshot, right-click, Open with › Google Docs. It creates a document containing the image plus everything it could read below. Slower, but better for a full page of text.

Files app preview on recent ChromeOS versions also recognises text in images directly.

When recognition gets it wrong#

Text recognition is very good and not perfect. The predictable failure modes:

• Lookalike characters. Capital I, lowercase l and the digit 1; capital O and zero; rn read as m. In a serial number or a password, that matters and you won't spot it.
• Small text. Below about 10 pixels of character height, accuracy falls sharply. Zoom in before capturing rather than after.
• Low contrast. Grey text on a grey background, or text over an image.
• Unusual fonts. Decorative and condensed faces, and handwriting.
• Coloured backgrounds reduce accuracy noticeably.

Always check anything that must be exact — codes, URLs, numbers, commands. Recognition errors in those are silent and expensive.

Getting better results#

• Zoom the content before capturing. Bigger characters in the original image is the single biggest factor.
• Capture a tight region around just the text.
• Increase contrast first where you can — switching an app to light mode often helps more than any post-processing.
• Capture columns separately to preserve reading order.

Pair it with a tight crop#

Recognition accuracy depends far more on how big the characters are in the image than on the software. Capture a tight region around just the text, and crop it before running recognition rather than after.

Why this is more useful than it sounds#

Text extraction quietly solves several everyday problems: copying an error code from a screenshot someone sent you, getting a URL out of a video frame, pulling numbers from a chart image into a spreadsheet, or copying text out of an application that deliberately prevents selection.