How AI Nail Biting Detection Works
Stop Biting uses MediaPipe and WebAssembly to detect nail biting in real time — entirely on your device. Setup takes under two minutes.
Why awareness is the bottleneck
Most nail biters catch fewer than half of their daily biting episodes through self-monitoring alone. The habit is automatic — it runs below the threshold of conscious awareness. Habit Reversal Training (the gold-standard treatment) identifies awareness training as its most critical component. Stop Biting automates that component: it catches the episodes you don't notice.
Getting started
- Open the app. Visit stopbiting.today in Chrome, Edge, or Firefox — or download the macOS/Windows desktop app.
- Grant camera access. Allow the app to use your webcam. The video is processed locally — nothing is ever transmitted.
- Position your webcam. Make sure your face and hands are visible in the camera view. The AI tracks hand-to-mouth movements.
- Work normally. The app runs in the background. When it detects nail biting, an audible alarm fires immediately.
- Perform your competing response. When the alarm fires, press both palms flat on your desk for 60 seconds — the physical incompatibility breaks the habit chain.
Technical specifications
- Detection model: Google MediaPipe Hand Landmarker (21 landmarks)
- Processing: WebAssembly — sandboxed, on-device, no network access
- Alarm latency: under 1 second from detection to alarm
- Incident logging: timestamped log stored locally — never transmitted
Common questions
What happens to my camera data?
Nothing. The video feed is processed frame-by-frame by the MediaPipe WASM binary running locally. No frames, no thumbnails, no data of any kind is sent to any server — you can verify this by watching your network traffic while the app runs.
Does it work on Mac, Windows, and Linux?
The web app works on any device with a modern browser and webcam — Mac, Windows, Linux, and Chromebook. Native desktop apps are available for macOS and Windows for system-tray background running.
Will it false-alarm when I'm eating or touching my face?
The detection model distinguishes sustained hand-to-mouth proximity from brief touches, and sensitivity can be adjusted in settings.