Live AI Fitness Coach
Real-time exercise coaching that connects pose tracking, rep and form analysis, and conversational AI.
Overview
Exercise coaching needs to respond to movement as it happens. A useful system must coordinate pose observations, repetition state, and conversation while handling temporary tracking loss.
The Python system integrates MediaPipe pose tracking with rep and form analysis, using Gemini Live tool calling to surface metrics and proactive coaching.
How it works
- MediaPipe supplies pose tracking to the Python system.
- Rep and form analysis maintain exercise-related state.
- Event-driven logic surfaces metrics and coaching through Gemini Live tool calling.
Technical implementation
- Events signal rep goals, severe form issues, and tracking loss or recovery.
- Tool calling connects exercise analysis to the conversational interface.
- Tracking-loss and recovery handling account for interruptions in pose visibility.
Pose estimation depends on camera placement, visibility, and movement. No clinical validation or quantified coaching accuracy is claimed.
Results
Integrated pose tracking, exercise analysis, event-driven state, and conversational coaching in a real-time Python system.
The supplied description identifies rep/form analysis and tracking loss/recovery behavior. Rep-count accuracy, latency measurements, and a test protocol were not supplied.