Tavus Real-Time AI Video Call Demo Shows an Avatar Calling Functions in Your Browser
TejasKumar_ · x · 2026-10-02
Engineer Tejas Kumar built a demo around Tavus's new real-time AI video call, holding a 3-minute call with "Priya" — a face that isn't a real person — while exposing the entire stack in the browser:
- Four-model pipeline: Sparrow-2 decides when you've actually finished speaking (tone + pauses), Raven-1 handles camera and screen-share vision, generation runs on tavus-gemma-4, and Tavus renders the face in real time.
- In-browser function calling: the avatar can call tools defined in the page — checking a speaking schedule, searching past talks, writing a React debounce hook live, reading code on your shared screen, recognizing objects held up to the camera, even switching the page theme — with every tool round trip visible.
- The page shows per-turn end-to-end latency, data-channel messages and full wire logs, exportable as JSON. No recording, 3-minute calls only.
A rare fully transparent public demo of the multi-model chain and tool calling behind AI video calls.
More from coding & agent
- Aviation's ASD-STE100 controlled language as an anti-AI-slop prompt hack, and where it fails — Paimaamu · 2026-10-02
- Pi Durable as statecharts: an interactive demo of crash-safe LLM agent harnesses — sloppenheimer · 2026-10-02
- Microsoft open-sources NVX, an ultra-light OpenVMM-based micro-VM sandbox for agentic workloads — unixterminal · 2026-10-02
- exe.dev's 'Run Fewer Agents': why task management isn't the fix for agent sprawl — charles_irl · 2026-10-02
- Building an agentic ML team: multi-agent pipeline with 40% token savings — kmeanskaran · 2026-10-02
- Claude Code creator: I don't prompt anymore, I write loops — a PM starter — aakashgupta · 2026-10-02