Emotion Copilot – Private Video Calls With an AI Participant
A solo build for the Open Agents Hackathon. A video conferencing platform with real-time emotion analysis, live transcription with speaker identification, AI meeting summaries, and a voice-enabled agent that joins the call, answers questions, and speaks back. The point was privacy: peer-to-peer WebRTC calls, session-only processing, zero data retention, deployed on the decentralized Akash Network with Venice for private inference.
Project Features
Zero Data Retention
Calls are peer-to-peer and analysis happens per session. Nothing is stored after the meeting ends.
Live Emotion Read
Confident, engaged, confused, or frustrated, tracked from facial cues in real time and shown alongside the transcript.
An Agent in the Room
The AI participant listens, answers questions out loud when addressed, and produces the summary when the call ends.
Decentralized Hosting
Runs on Akash with Venice for inference, so neither the video nor the transcript touches a centralized AI provider.
Project Info
02/2026
Sole Developer
Solo Project
Hackathon Submission
Technologies Used
- • WebRTC
- • Browser-side face and emotion detection
- • Speech-to-text with speaker diarization
- • Venice AI
- • Akash Network
Watch It Run
Emotion Copilot demo
A live call with real-time emotion tags, running transcription, and the AI participant answering out loud.
Project Gallery
Interested in This Project?
I'm always excited to discuss my projects and share insights about the development process. Feel free to reach out if you'd like to know more!