Copy and translate text on screen
Updated October 24, 2026
OhMyShot treats the text in your screen as text. It reads it on your Mac with Apple’s Vision and its own post-processing, and nothing leaves the Mac unless you set up a cloud model with your own key.
Capture Text
Press ⇧⌘2 and drag over the words you want. The text in the area is copied to the clipboard.
- Settings › Tools › Image line breaks chooses what happens to line breaks: Keep Line Breaks (the default) keeps the rows as the image has them, and Join Lines merges the wrapped lines of each paragraph. Code, table rows, and formulas keep their structure either way.
- Capture Text, Joining Lines always joins, whatever the setting. It has no shortcut until you give it one in Settings › Shortcuts.
The Recognized Text window
The window shows a picture’s text beside the picture. Open it with Recognize Text on a capture’s Quick Access card, or with Recognize Text in Image… and Recognize Text in Clipboard Image in the menu bar menu. Beside the text it lists what it found in the picture: tables, QR codes and barcodes, and the links, email addresses, phone numbers, dates, and addresses in the text.
Its Line Breaks menu switches between Keep Line Breaks and Join Lines for that window only; Copy copies what the window shows. Its Translate toggle translates the text, also drawn over the picture.
Translate Area
Press ⇧⌘1 and drag over text in another language. The translation is painted over the original in place: each line is covered with its own background and set in its own ink.
- It uses Apple’s on-device Translation. A language pair already downloaded works at once; for a pair macOS can download, the system shows its own sheet first.
- Simplified and Traditional Chinese, which Apple’s Translation doesn’t translate between, are converted character by character.
- Code, figures such as “5 GB”, links, file names, and commands are left as they are.
Model packs
A model pack reads what macOS reads poorly or not at all. Packs are optional, downloaded once in Settings › Text › Model packs, and kept on your Mac; nothing you capture leaves it. They are read in a helper process that exits when it is done.
| Pack | Size | What it reads | Model |
|---|---|---|---|
| Chinese and Japanese Text | 15.8 MB | Chinese and Japanese more accurately than macOS alone, in the Recognized Text window and in Capture History’s search | PP-OCRv6 small, by PaddlePaddle |
| Math Formulas | 219 MB | Each formula as LaTeX in recognized text; Capture Formula copies the formula you select | PP-FormulaNet_plus-L and PP-DocLayoutV3, by PaddlePaddle |
| Tables and Charts | 1.12 GB | Each table and chart as rows you can paste into a spreadsheet; reading one takes about 2.6 GB of memory | PaddleOCR-VL-1.6 and PP-DocLayoutV3, by PaddlePaddle |
Every pack comes from Hugging Face under the Apache-2.0 license. The first download of a pack opens a sheet first, with what it is, its size, where it comes from, links to its license and the original model, and the license’s full text; nothing downloads until you agree. Each file is kept only if its size and SHA-256 match the ones built into the app, and a download that stops resumes where it left off.
Cloud reading, with your own key
Where your Mac reads a picture poorly (handwriting, dense tables, complex layouts), Settings › Text › Cloud reading lets a model you have an account with read it, such as Claude, GPT, or Qwen. Nothing is sent until you choose Read with… in the Recognized Text window, or turn on Capture Text or every picture in that window in Settings; the picture then goes from your Mac straight to the service, which may charge for it. Keys stay in the Keychain.
URL commands, Shortcuts actions, and Capture History’s search always read on your Mac.
Text from apps
With Settings › Text › Text from apps on (it is off by default, and turning it on asks for Accessibility), Capture Text and the Recognized Text window ask the app under the selection for the text it shows, and use the app’s own characters for the lines that read the same, so code, names, and numbers come out exactly as the app has them. Password fields are passed over. Apps that don’t share their text, such as Chrome and most apps built with Electron, are read from the picture alone, and so are pictures from files, the clipboard, pins, and Capture History.
Searching Capture History
Capture History can be searched by the text in your captures, Chinese included. With the Chinese and Japanese Text pack installed, the search reads Chinese and Japanese with it too.