Atmosphere is a AI-powered music generation system that creates personalized soundtracks by analyzing your visual surroundings in real-time. Built for HackMIT 2025 as a collaborative group project, this system combines Mentra smart glasses with cutting-edge AI services to deliver seamless, adaptive music experiences that respond to your environment.
The system features a sophisticated pipeline: Mentra glasses capture images every 3 seconds, which are processed by Claude Vision API for scene analysis. The backend, built with Express.js and TypeScript, orchestrates AI services including Suno AI for music generation. Real-time communication is handled through WebSocket connections, while Redis manages caching and session storage. The companion app, developed in React, provides user interface and music history tracking.
Team Members: Srinivas (Full Stack & Hardware Integration), Ayat (App Development & Hardware Integration), Zayyan (Vision Analysis & Backend Integration), Eric (Presentation & Suno Integration)
My Contributions: I led the complete frontend development, building the React companion app from scratch with real-time status monitoring and music history tracking. I assisted with hardware integration for the Mentra glasses and implemented API calls for seamless data streaming between the glasses and backend services. I also designed the UI/UX interface and helped optimize the communication pipeline.
Frontend: React, TypeScript, HTML5/JavaScript
Backend: Node.js, Express.js, Bun runtime
Communication: Socket.IO, WebRTC
AI Services: Claude Vision API, Suno AI
Infrastructure: Redis, Docker, Sharp (image processing)
Hardware: Mentra Smart Glasses
Repository: github.com/IBS27/atmosphere