On-Device Privacy Check
Background Motion & Privacy Leak Scanner
This webcam background privacy check exists because it is easy to forget what is visible behind you — a whiteboard with passwords, a screen with an open document, a hallway where people walk by. This highlights high-motion zones and periodically runs a real, fully client-side OCR pass to flag any readable text in frame. Nothing is uploaded — the OCR model runs in your browser via WebAssembly.
Legend
Reading the Overlay
Readout
Scan Results
Click "Start Scan" — the motion map updates live, and an OCR pass runs every few seconds.
Motion
- Persistent Motion Zones:—
- Busiest Zone:—
Text Detection (OCR)
- OCR Status:Not started
- Last Scan:—
- Text Regions Found:—
OCR quality depends heavily on lighting, angle, and text size — this flags likely text, it doesn't guarantee catching everything readable in frame. Treat a clean scan as a helpful check, not a guarantee.
Notes
What your camera reveals that you stop noticing
People become blind to their own backgrounds within days of setting up a desk. A whiteboard with a client name on it, a delivery label with your home address, a monitor showing an email subject line, a prescription box on a shelf — all of it sits in frame on every call, and none of it registers because you stopped seeing that wall months ago.
The risk is not usually dramatic. It is ordinary information disclosed to people who happen to be on the call, sometimes recorded, occasionally shared onward. A background scan is worth thirty seconds before a call with anyone outside your organisation.
How the two detectors work
Motion mapping. The frame is divided into a grid and each cell tracks how much it changes between frames. Cells that keep changing over a sustained period are flagged. This catches the things you cannot control — a doorway where someone walks past, a window onto a street, a television left on behind you.
Text recognition. An optical character recognition engine runs periodically over the frame and draws a box around anything it can read. If software can read it from your webcam feed, so can a person watching the call, and so can anyone who receives a recording.
Both run entirely on your device. The OCR engine is downloaded to your browser and executes locally, which is the only responsible way to build a tool whose whole purpose is finding sensitive information.
Acting on the results
Limits worth knowing
OCR reads clear, reasonably sized, front-facing text. It will miss handwriting, small print, text at a steep angle, and anything low in contrast — so a clean scan is not proof that nothing is legible. Human eyes are better than software at spotting a partially obscured word, and a viewer can pause a recording and zoom in.
Use this to catch what you have stopped noticing, then look at your own background properly. The tool is a prompt to check, not a certificate that you are safe.
Reference
What Gets Detected
| Condition | Detected? |
|---|---|
| Clear, front-facing printed text | Reliably |
| Small, angled, or low-contrast text | Inconsistently |
| Handwriting | Rarely |
| A doorway, window, or screen with sustained movement | Reliably |
| Brief or one-off movement | May not persist long enough to flag |
This is not a scored result — a clean scan means nothing was detected under these real constraints, not a certified "no risk" rating.
Step-by-Step
How to Run the Scan
- Click "Start" and allow camera access when prompted, then wait a moment for the OCR engine to load.
- Sit in your normal call position so the scan sees exactly what others would.
- Watch for highlighted text boxes and motion zones as they appear over a few seconds of scanning.
- Address anything flagged, then look at your background yourself — OCR misses handwriting and small print that your own eyes will catch.
Common Questions