AI-powered fitness tracker that runs entirely in your browser — no installs, no servers. Uses TensorFlow.js and ElevenLabs AI Voice to track your form, count reps, and give real-time personalized voice feedback with different coach personalities.

👉 Live Demo

• 🧠 Real-time pose detection with MoveNet (TensorFlow.js)
• 🗣️ AI Voice Feedback powered by ElevenLabs
• 🔢 Automatic rep counter with form quality tracking
• 📊 Personal Best Tracking with Supabase database
• 🏆 New Record Celebrations with visual and audio effects

• 🤸 Jumping Jacks Tracker - Tracks arm elevation and leg spread
• 🦵 Squat Tracker - Monitors knee angle and back posture

Choose your AI coach style:

• 🎖️ Military Sergeant - Strict & commanding
• ✨ Pop-up Girl - Energetic & bubbly
• 💪 Gym Bro - Casual & supportive

• 📐 Real-time angle measurements (arms, legs, body alignment)
• ⚠️ Form validation with instant corrective feedback
• 🎯 Confidence tracking for pose detection accuracy
• 🔒 100% client-side processing — camera data never leaves your device
• 📱 Responsive design for desktop and mobile
• 🎵 Special audio effects at milestone reps

Frontend:

• HTML5, CSS3, JavaScript (ES6+)
• TensorFlow.js with MoveNet (Pose Detection)
• Web Speech API (fallback voice)

AI & Services:

• ElevenLabs API (Premium AI voice synthesis)
• Supabase (Database for personal bests)

Hosting:

• GitHub Pages / Vite

1. Open the live demo link
2. Choose your coach personality (Military, Pop-up Girl, or Gym Bro)
3. Select an exercise (Jumping Jacks or Squats)
4. Allow camera access when prompted
5. Get into position and start exercising
6. Hear real-time AI voice guidance and see:
   • Rep count
   • Form metrics (angles)
   • Personal best comparison
   • Instant feedback on your form

• Node.js (optional, for local server)
• Modern browser with webcam support

git clone https://github.com/<your-username>/FormAI.git
cd FormAI
python -m http.server 8000
# or use any local web server

Visit: http://localhost:8000

Create a .env file with:

VITE_SUPABASE_URL=your_supabase_url
VITE_SUPABASE_ANON_KEY=your_supabase_key

FormAI/
├── index.html              # Main app layout
├── style.css               # Styling and animations
├── script.js               # Main app logic
├── audioManager.js         # ElevenLabs integration & audio handling
├── exerciseDetectors.js    # Jumping Jacks & Squat detection algorithms
├── formValidators.js       # Real-time form validation
├── stateMachine.js         # Exercise state management
├── poseUtils.js            # Pose detection utilities
├── feedbackManager.js      # Voice feedback coordination
├── personalityMessages.js  # Coach personality dialogue
├── personalBestManager.js  # Personal records tracking
├── supabaseClient.js       # Database connection
└── supabase/
    └── migrations/         # Database schema

1. Camera Feed → TensorFlow.js captures video frames
2. MoveNet Model → Extracts 17 body keypoints
3. Angle Calculation → Computes joint angles in real-time
4. State Machine → Tracks exercise phases (ready → descending → bottom → ascending → complete)

• Tracks arm elevation angle (shoulder-to-wrist)
• Measures leg spread distance (ankle separation)
• Rep counted when: arms up + legs spread → arms down + legs together

• Real-time checks for proper form
• Persistence tracking prevents false warnings
• Priority system for critical vs warning-level feedback

• ElevenLabs generates natural AI voice
• Fallback to Web Speech API if needed
• Context-aware messages based on exercise state
• Debounced playback prevents audio conflicts

✅ Enhanced side profile detection - Lower confidence thresholds allow tracking when turning sideways ✅ Audio error handling - Fixed play/pause conflicts and TensorFlow texture errors ✅ Jumping Jacks support - Replaced push-ups with full jumping jacks tracking ✅ Robust error handling - Graceful degradation for audio and video issues

• Angle smoothing with rolling buffers
• Debounced audio playback
• Video dimension validation
• Cached TTS audio responses

• No data collection - All processing happens locally in your browser
• No video upload - Camera feed is processed client-side only
• Optional cloud features - Personal bests stored in Supabase (can be disabled)
• Secure API keys - Environment variables for sensitive credentials

• Clean, modern design with glassmorphism effects
• Real-time visual feedback with skeleton overlay
• Dynamic status indicators with color-coded states
• Animated transitions between screens
• New record celebrations with overlay animations
• Metric displays showing angles and confidence scores

Camera not working?

• Ensure you've granted camera permissions
• Check if another app is using the camera
• Try refreshing the page

No voice feedback?

• Check browser audio permissions
• Verify ElevenLabs API key is configured
• Falls back to Web Speech API automatically

Pose not detected?

• Ensure good lighting
• Position yourself fully in frame
• Move away from the camera for full body visibility

• Add more exercises (planks, burpees, lunges)
• Workout session tracking & analytics
• Multi-user support with profiles
• Mobile app version
• Social sharing & challenges
• Custom coaching plans
• Integration with fitness wearables

Contributions welcome! Please feel free to submit a Pull Request.

MIT License - Feel free to use this for your own projects!

Built with ❤️ for fitness enthusiasts — AI-powered form tracking with no backend required.