A quick web app to check if there's something stuck in your teeth after a meal. Just smile at your camera and get an instant yes/no answer!
- Quick Launch - PWA-enabled, installable on your home screen for instant access
- Front-Facing Camera - Automatically uses selfie camera
- AI-Powered Analysis - Uses on-device machine learning to analyze your smile
- Privacy First - All processing happens in your browser, no images sent to servers
- Pretty UI - Modern, mobile-first design with animated results
- Works Offline - After first load, works without internet
- React 19 + TypeScript - Modern UI framework
- Vite 7 - Fast build tool with PWA support
- Tailwind CSS 4 - Utility-first styling
- Transformers.js - On-device ML inference (Hugging Face)
- vite-plugin-pwa - Service worker and manifest generation
- Node.js 18+
- npm or yarn
# Install dependencies
npm install
# Start development server
npm run dev
# Build for production
npm run build
# Preview production build
npm run previewnpm run generate-pwa-assets- Camera Capture - Uses the Web MediaDevices API to access the front-facing camera
- Photo Capture - Takes a snapshot when you tap the capture button
- AI Analysis - Uses a pre-trained image captioning model (vit-gpt2) to describe what it sees
- Result Parsing - Analyzes the model output for indicators of food debris vs clean teeth
- Display Result - Shows a clear YES/NO with confidence level
- Chrome/Edge 90+
- Safari 15+
- Firefox 90+
Best experience on mobile devices (iPhone/Android) where you can install as PWA.
# Run linter
npm run lint
# Type check
npm run buildThis is an MVP with known limitations:
- The AI model is a general image captioning model, not specifically trained for teeth debris detection
- Accuracy depends on lighting conditions and image quality
- False positives/negatives are possible - when in doubt, check a mirror!
MIT