Feature

Read the text in any screenshot, on your computer

Select words in a capture like normal text, copy them, translate them or ask Claude about them. The recognition runs locally and nothing is uploaded for it.

How it works

Text selection starts off for every capture, so clicking around never selects words by accident.

  1. CapturePress CtrlShift1 and select a region. SnapPrompt recognizes the text in it on your computer.
  2. Turn on text selectionPress AltS or choose Select text in the capture tool strip. Press it again to turn selection off.
  3. Select wordsSelect words in the image like normal text. An action bar appears above the selection, or below it when there is no room.
  4. Copy, ask, translate or look upCopy the text, ask Claude about it, translate it or look it up. Text you select is sent as a quote.

The engine on each system

  • Windows: Windows OCR is built into Windows and needs no download. Or chooseXiaomi-OCR-0, a local model of about 850 MB that you download from Settings. Windows OCR still shows text right away, and the model refines it in the background.
  • Mac: the text recognition built into macOS (Apple Vision). It needs no download.
  • Linux: Tesseract, installed from your distribution's packages, withtesseract-ocr-vie for Vietnamese on Ubuntu and Debian. Without it captures still work, but text may not be recognized, and Settings shows the install command.

Xiaomi-OCR-0 is available on Windows only for now. Turn OCR off, or pick the engine on Windows, in Settings under Capture and privacy. When the Windows language pack for Vietnamese is missing, the app tells you, because diacritics may be lost.

In the editor

Open the full image editor with AltE (on a Mac: Edit capture on the answer card). The Recognized text panel in the top bar lets you copy all the text, or select part of it to ask Claude about only that part.

The same recognized text powers auto-redact, which covers emails, phone numbers, card numbers, API keys and tokens it finds.

Frequently asked questions

Is my screenshot uploaded to read its text?

No. SnapPrompt recognizes the text in a capture on your computer, and nothing is uploaded for OCR. If you then ask Claude about text you selected, that selected text is sent as a quote, because you asked Claude about it.

Which OCR engines does SnapPrompt use?

It depends on your system. On Windows: Windows OCR, which is built in, or the local Xiaomi-OCR-0 model, about 850 MB, which you download from Settings. With Xiaomi-OCR-0, Windows OCR shows text right away and the model refines it in the background. On a Mac: the text recognition built into macOS (Apple Vision). On Linux: Tesseract, installed from the packages of your distribution. Xiaomi-OCR-0 is available on Windows only for now.

Does it read Vietnamese?

Yes. On Windows it needs the Windows language pack for Vietnamese, and SnapPrompt warns you when the pack is missing, because diacritics may be lost without it. On a Mac, SnapPrompt uses the text recognition built into macOS, so there is no language pack to install; it reads Vietnamese on the macOS versions whose text recognition supports it. On Linux it needs the Vietnamese data for Tesseract (tesseract-ocr-vie on Ubuntu and Debian), and Settings shows the install command when it is missing.

Can I turn OCR off?

Yes. OCR on or off, and the OCR engine, are in Settings under Capture and privacy. An organization policy can also choose the engine.

How do I copy text from a screenshot?

Press Alt+S, or choose Select text in the capture tool strip, then select words in the image like normal text and copy them. In the editor, Recognized text has a Copy all button.

Try SnapPrompt free for 3 days

Windows, Mac and Linux. The macOS and Linux builds are coming soon. After the trial, $19 + applicable tax once, with 1 year of updates included. You use your own Claude account.