AStack Demo

Client-Side Rendering Demo

AI Avatar

Avatar Type:

Controls & Status

Status:Disconnected

Event Log

No events yet...

How It Works

This demo showcases AStack's client-side rendering pipeline for real-time AI avatar interactions.

Client-Side Rendering

  • • Avatar rendered in-browser (VRM/TalkingHead)
  • • Audio playback with blendshape sync
  • • WebSocket for data transfer (not WebRTC)
  • • 52 ARKit blendshapes for facial animation
  • • Low bandwidth - only audio + blendshape data

AI Pipeline

  • • Real-time speech recognition (ASR)
  • • Large language model responses (LLM)
  • • Text-to-speech synthesis (TTS)
  • • Audio-to-face blendshape generation
  • • Sub-second response latency target

Demo Flow

  1. 1. Click "Connect to AStack" to establish WebSocket connection
  2. 2. Select AI providers (ASR, LLM, TTS)
  3. 3. Click "Start Call" and allow microphone access
  4. 4. Speak to the AI - watch the avatar respond
  5. 5. Use text input for silent messaging
  6. 6. Switch between VRM and TalkingHead avatars

Client-Side Rendering Enabled! This demo uses the @astack/client-sdk with client-side avatar rendering. Audio and blendshapes are streamed via WebSocket, while the 3D avatar is rendered entirely in your browser.