JNKS Yoga

AI-powered yoga form analyzer. Record your practice via webcam or upload a video, and get instant feedback on your alignment, balance, flexibility, and form. Uses Gemini AI vision for analysis — no local ML models required.

How It Works

1. Record your practice using a webcam (live analysis) or upload a video file
2. Analyze your form with Gemini AI vision (video sent to API)
3. Score across 4 categories with detailed per-section feedback
4. Coach with real-time voice coaching via Gemini Live

Tech Stack

Frontend

• Next.js 16 (React 19) with TypeScript
• Tailwind CSS 4, shadcn/ui components
• Framer Motion, GSAP, Anime.js (animations)

AI Pipeline

• Gemini Pro (video analysis per clip)
• Gemini Flash (structured scoring, coaching distillation)
• Gemini TTS (text-to-speech coaching)
• Gemini Live (real-time voice coaching session)
• FastAPI + WebSocket backend

Prerequisites

• Node.js 18+
• Python 3.8+
• ngrok (tunnels the backend so your phone can connect)

  brew install ngrok
  ngrok config add-authtoken YOUR_TOKEN

No GPU or local ML models needed — all analysis runs through the Gemini API.

Getting Started

1. Clone and install

git clone <repo-url>
cd JNKS-Yoga
npm install

2. Set up the Python backend

# Option A: create fresh venv
python3 -m venv venv
source venv/bin/activate
pip install -r pipeline/requirements.txt

# Option B: reuse JNKS venv (if you have the JNKS repo)
ln -s ../JNKS/venv venv

3. Configure environment

Create a .env file in the project root:

GEMINI_API_KEY=your_gemini_api_key

4. Run

./start.sh

This will:

• Start an ngrok tunnel on port 8001
• Build the frontend with the ngrok URL baked in
• Start the FastAPI backend
• Print the URL to open on your phone

Access:

• Phone/remote: the ngrok URL printed by the script
• Local: http://localhost:8001

Press Ctrl+C to stop everything.

Manual start (without ngrok)

source venv/bin/activate
npm run build
python3 -m pipeline.server

Then open http://localhost:8001.

Project Structure

src/
  app/              # Next.js pages (landing, analyze, results)
  components/
    analyze/        # Webcam feed, video upload, analyze button
    landing/        # Hero, features, how-it-works sections
    results/        # Score display, breakdown, tips
    layout/         # Navbar, theme, animated background
    ui/             # Reusable UI components
  lib/              # Utilities, API helpers
  types/            # TypeScript interfaces

pipeline/           # Python AI pipeline
  server.py         # FastAPI server (REST + WebSocket)
  gemini_vision.py  # Gemini video analysis per clip
  scoring.py        # Score computation from Gemini results
  llm.py            # Gemini text generation, TTS
  voice.py          # Gemini Live real-time voice coaching
  video.py          # Video conversion and splitting (ffmpeg)
  storage.py        # Simple JSON storage

start.sh            # One-command startup (ngrok + build + backend)

Scoring Categories

Category	Weight	Description
Alignment	30%	Spine neutrality, joint stacking, head/neck/pelvis position
Balance	25%	Stability, weight distribution, grounding
Flexibility	25%	Range of motion, depth, joint mobility
Form	20%	Overall pose quality, symmetry, muscle engagement

Each category is scored 0-100 using hybrid scoring: Gemini vision analysis blended with pose metrics. The coaching advice targets your weakest area with specific, actionable tips.

Key Differences from JNKS (Basketball)

	JNKS (Basketball)	JNKS Yoga
Port	8000	8001
Local ML	MediaPipe + YOLOv8	None (cloud-only)
Python deps	10 packages (heavy)	7 packages (lightweight)
Analysis	Pose comparison to pro references	Gemini vision scoring
Voice coaching	TTS playback	Real-time Gemini Live conversation