MediVoice

• This image is currently processing. Please be patient...
  

Inspiration

Accessing medical guidance is often delayed, expensive, or inaccessible.

What it does

MediVoice is a real-time medical consultation assistant: describe symptoms via voice, show physical symptoms via camera, upload medical documents, get immediate voice-guided triage.

How we built it

Frontend: HTML/CSS/JS + Web Audio API. Backend: Node.js/Express on Google Cloud Run (WebSocket proxy). AI: Gemini 2.0 Flash Live API for real-time audio + vision.

Challenges we ran into

PCM audio encoding, latency optimization, simultaneous camera + voice through one WebSocket.

What we learned

Gemini Live API interruption is natural. Raw PCM = lower latency. Medical AI needs careful prompt engineering.

What's next

Multi-language support, health records integration, specialist referrals.

Built With

• express.js
• google-cloud-run
• node.js
• web-audio-api
• websocket