• 

  Choose which body part you want to rehabilitate

• 

  Choose intensity and one of the real time generated workouts

• 

  Realtime video generated using Veo which provides a demo for the exercise

• 

  Gemini 3 api guides you throughout the workout by monitoring your pose. You can also ask doubts with real time audio

Inspiration:

Traditional physiotherapy is repetitive, boring, and often inaccessible, especially for patients recovering from neurological or orthopedic conditions. Many patients stop rehab early due to lack of motivation, supervision, or personalized guidance. We wanted to reimagine rehabilitation as something engaging, intelligent, and adaptive, using generative AI to act like a personal physio coach available anytime, anywhere.

What it does:

Rehabify is an AI-powered rehabilitation platform built using the Gemini 3 API. Users select a body part they want to rehabilitate, and Rehabify instantly generates a live exercise demo video using Veo, tailored to that body part. As users perform the exercise, Gemini Live monitors their movements in real time, provides instant corrective feedback, motivates them with voice/text prompts, and increases their score for every correct repetition, turning recovery into a game.

How we built it:

• Gemini 3 API for real-time reasoning, feedback generation, and live exercise monitoring
• Veo to generate realistic, AI-powered exercise demo videos on demand
• Computer vision & pose analysis to track user movements during exercises
• Gamification layer to calculate accuracy and score per repetition
• Web-based frontend for accessibility and real-time interaction

Challenges we ran into:

• Designing real-time feedback that is helpful, not overwhelming
• Synchronizing live AI monitoring with video demos
• Handling variations in user movement, camera angles, and lighting
• Balancing medical seriousness with fun gamification
• Making AI feedback feel supportive rather than robotic

Accomplishments that we’re proud of:

• Successfully integrating Gemini Live + Veo into a real-time rehab workflow
• Creating a fully AI-driven physio experience without pre-recorded videos
• Turning rehabilitation into a motivating, score-based game
• Building a system that is scalable across multiple body parts and conditions

What we learned:

• Generative AI can act as a real-time coach, not just a chatbot
• Visual feedback + motivation dramatically improves user engagement
• Designing for healthcare requires empathy, clarity, and trust
• Gemini’s multimodal reasoning unlocks entirely new interaction patterns

What’s next for Rehabify: • Personalized rehab plans based on injury or condition history • Progress tracking dashboards for patients and doctors • Voice-based coaching and multilingual accessibility • Adaptive difficulty using reinforcement learning • Clinical validation with physiotherapists and hospitals

Built With

• api
• css
• express.js
• express.js-?-platform:-web-(browser-based-for-accessibility)-?-cloud-&-ai-services:-google-ai-studio-/-google-cloud-(gemini-&-veo)-?-other:-real-time-scoring-&-gamification-engine
• gemini
• gemini-live
• gemini3
• google-ai-studio
• google-cloud
• html
• javascript
• languages:-javascript
• live-feedback)
• mediapipe
• node.js
• python
• python-?-frontend:-react.js
• react
• tailwind-css-?-ai-&-apis:-gemini-3-api-(multimodal-reasoning
• veo
• veo-(ai-video-generation)-?-computer-vision:-pose-estimation-&-motion-tracking-(web-based-cv-pipeline)-?-backend:-node.js
• veo-video-converter