Extract & redact screen text with local Safe OCR
When text is trapped inside an uncopyable error dialog, RDP window, VM console, or video frame, press Ctrl+Shift+A to open Safe Capture. SafeRelay runs local OCR on your machine, applies local DLP protection, and lets you review the protected text before copying, pasting, or saving.
Select Region & Choose Action
Press Ctrl+Shift+A, drag across the uncopyable dialog or VM region, and select Safe OCR + Redact.
Recognize & Transform
Text is recognized locally on Windows and immediately scanned by the local DLP engine.
Inspect & Choose Output Action
OCR Accuracy & Mandatory Human Review
OCR is best-effort. Review protected output before using it. Zooming in can improve recognition accuracy. Because optical character recognition depends on font rendering, display scaling, contrast, and kerning, an OCR misread (such as confusing 0 and O or splitting a token across lines) can prevent a pattern detector from matching part of a secret. Always inspect the review window before clicking Copy, Paste & close, or Save.
How Safe OCR Redaction works with Ctrl+Shift+A
Press Ctrl+Shift+A
Trigger Safe Capture from anywhere on Windows and drag a selection rectangle around the dialog, VM console, or image containing text.
Select Safe OCR + Redact
From the Safe Capture action palette, choose Safe OCR + Redact (or choose Extract text locally when you only need local OCR extraction).
Review Protected Output
SafeRelay displays the recognized text after local DLP transformation. Verify that all sensitive characters were properly recognized and redacted.
Copy, Paste & Close, or Save
Once verified, choose Copy, Paste & close, or Save to use the sanitized text in your destination workflow.
Inside the Ctrl+Shift+A Safe Capture Palette
Safe Capture consolidates screen-region workflows under one shortcut (Ctrl+Shift+A) with clear boundaries between local-only actions and optional AI actions:
- Safe OCR + Redact (100% Local): Runs local OCR on the captured screen region, applies local DLP redaction and Custom Protected Terms, and opens the review window.
- Extract text locally (100% Local): Extracts text from the screen region on-device without sending image data to any server.
- Optional AI commands: Only invoked if you explicitly configure and select an AI command in the palette. Local OCR and local DLP never require a cloud connection.
Tips for Higher OCR Recognition Accuracy
When capturing dense terminal fonts or small UI dialogs:
- Zoom in before capturing: Increasing the zoom level in your browser, PDF viewer, or RDP window sharpens character edges and improves recognition accuracy.
- Crop tightly around the text block: Avoid including decorative icons, window borders, or overlapping scrollbars in the selection rectangle.
-
Prefer selectable text when available: If the text can be highlighted directly, use
Ctrl+Shift+8(browser) orCtrl+Shift+9(desktop) for exact character-level fidelity.