Like having a physiotherapist in your living room
Demo · Features · How It Works · Tech Stack · Quick Start
After spending $8K on physio bills, waiting 3 weeks for my first appointment, and only getting seen once every 2 weeks for a herniated disc, I realized the broken part of physical therapy isn't the clinic—it's what happens at home.
| The Reality of Physical Therapy |
|---|
| Only 35% of patients complete their prescribed exercises |
| Patients are unsure if they're doing exercises correctly |
| Zero supervision between clinic visits |
| PTs have no visibility into home exercise quality |
Rehabify uses computer vision + voice AI to provide real-time form correction using just a webcam. Perform your exercises at home and get instant feedback like "Keep your back straight" or "Lower those hips"—transforming PT from opaque guesswork into data-driven care.
| Poor Form Detection | Mid-Rep Coaching | Good Form Achieved |
|---|---|---|
 |
 |
 |
| 15% - "Sit back more, keep knees behind toes" | 83% - "Squat a bit deeper" | 92% - "Good depth, hold it!" |
- Real-Time Form Correction — AI watches your movement and provides instant voice feedback
- Personalized Exercise Plans — Guided assessment generates a tailored 12-week program using Gemini AI
- Progress Tracking — See your completion rates, form scores, and improvement over time
- Privacy-First — Video never leaves your device; all computer vision runs locally in your browser
- Remote Monitoring Dashboard — View patient completion rates, form quality scores, and problem areas
- Data-Driven Insights — Transform subjective patient reports into objective movement data
- Safety Guardrails — System detects serious symptoms and redirects patients to in-person care
- Medical Safety Stops — Serious symptoms (numbness, bowel issues, severe pain) trigger immediate redirection to professional care
- HIPAA-Ready Architecture — Built with healthcare data privacy in mind
- Canadian Healthcare Compliant — Designed for integration with Canadian healthcare systems
┌─────────────┐ ┌──────────────┐ ┌─────────────────┐ ┌─────────────┐
│ Webcam │────▶│ MediaPipe │────▶│ Form Detection │────▶│ Voice AI │
│ (30fps) │ │ Pose │ │ Engine │ │ Coach │
└─────────────┘ └──────────────┘ └─────────────────┘ └─────────────┘
│ │ │
33 body landmarks Joint angle "Lift those
(runs locally in analysis & hips up!"
WebAssembly) rule-based (< 1 sec
validation latency)
- Capture — Webcam streams video at 30fps
- Detect — MediaPipe Pose extracts 33 body landmarks in real-time (runs entirely in browser via WebAssembly)
- Analyze — Custom algorithms evaluate joint angles, body alignment, and movement patterns
- Feedback — Voice AI delivers natural coaching feedback in under 1 second
| Layer | Technology | Purpose |
|---|---|---|
| Frontend | Next.js 15, React 19, Tailwind CSS | "Wellness sanctuary" design system |
| Computer Vision | MediaPipe Pose (WASM + Web Workers) | Real-time skeleton tracking |
| Voice AI | Vapi (Deepgram STT + Gemini + ElevenLabs TTS) | Natural coaching feedback |
| Database | Neon PostgreSQL + Drizzle ORM | User data & exercise tracking |
| AI/ML | Google Gemini API | Personalized plan generation |
| Deployment | Vercel | Edge-optimized hosting |
- Node.js 18+
- pnpm
- PostgreSQL database (we use Neon)
# Clone the repository
git clone https://github.com/yourusername/rehabify.git
cd rehabify
# Install dependencies
pnpm install
# Set up environment variables
cp .env.example .env.local
# Add your API keys (Neon, Vapi, Gemini)
# Initialize database
npx drizzle-kit push
# Start development server
pnpm devVisit http://localhost:3000 to see the app.
pnpm test # Run all tests
pnpm test:watch # Watch mode
pnpm test:e2e # Playwright E2E tests
pnpm test:coverage # Coverage report| Tier | Exercises | Capabilities |
|---|---|---|
| Tier 1 | Cat-Camel, Cobra, Dead Bug, Glute Bridge, Bodyweight Squat | Full AI form detection with real-time feedback |
| Tier 2 | 47+ additional exercises | Voice-guided with reference videos |
All exercises sourced from licensed PT resources and validated for safety.
rehabify/
├── plan/ # Detailed specs and architecture docs
│ ├── 00-architecture/ # System design, data flow
│ ├── 01-backend/ # Database, API routes
│ ├── 02-frontend/ # Components, wireframes
│ ├── 03-vision/ # MediaPipe, form detection rules
│ ├── 04-voice-ai/ # Vapi integration, prompts
│ └── 05-contingencies/ # Scope reduction plans
│
├── src/
│ ├── app/ # Next.js App Router pages
│ ├── components/ # React components
│ ├── lib/ # Vision, voice, database utilities
│ │ ├── mediapipe/ # Pose detection & form analysis
│ │ ├── vapi/ # Voice AI integration
│ │ └── gemini/ # Plan generation
│ └── stores/ # Zustand state management
│
└── public/
├── videos/ # Exercise demo videos
└── exercise-images/ # Exercise reference images
- MediaPipe Pose is incredibly powerful but requires careful tuning for different body types
- Voice AI has come far—Gemini can give nuanced coaching feedback that feels human
- Privacy-first architecture is possible AND performant with modern browser APIs
- Next.js 15 App Router + React Server Components are perfect for healthcare apps with mixed public/private data
- Expand from 5 to 52 exercises (full PT library)
- Mobile app (React Native) for better camera angles
- PT onboarding flow for custom exercise plans
- Machine learning for personalized form validation
- Progress prediction: "73% likely to recover in 4 weeks"
- Insurance integration (CPT code 98975)
- Clinical trials with UBC PT department
Physical therapy shouldn't require proximity to an expensive clinic. It should be something you can do confidently at home, with real-time guidance and professional oversight.
Rehabify is the infrastructure to make that happen.