Links
Overview
Eric Chen and I explored combining face-video analysis with a conversation about how a user feels. The goal was to make physiological data easier to interpret without a dedicated wearable.
Process
The stack combines a Next.js dashboard, Python/FastAPI, MongoDB, an OpenCV rPPG pipeline, Gemini, and ElevenLabs. Tracking a forehead region helps reduce motion noise, while an initial conversation fills the measurement buffer.
Result
Built a hackathon prototype integrating video, voice, and conversational context. Its repository documents timeouts, model fallbacks, and camera-status handling.
Gallery
