Real-time Audio & Video (Live API)
Real-time Audio & Video (Live API)
Gemini Live API provides bidirectional streaming across audio and visual modalities. AMC WebUI delivers a rich UI client for real-time interaction.
Core Interaction Modes
1. Natural Voice Streaming
- Leverages the browser’s native AudioWorklet API to record 16kHz PCM audio and stream directly over WebSockets.
- Supports instant interruption: speaking naturally interrupts model speech immediately.
- Real-time frequency waveform visualizer for acoustic clarity.
2. Camera Vision
- Stream frames directly from your webcam.
- The model continuously analyzes visual objects, diagrams, text, or hardware components in real time.
3. Screen Sharing
- Share application windows, Chrome tabs, or your full desktop.
- Ideal for live coding reviews, troubleshooting system errors, or navigating software workflows.
Communication Modes & Proxies
| Deployment Mode | Live Connection | Best For | Details |
|---|---|---|---|
| Docker Compose | Full WebSocket Proxy (/api/live) | Self-hosting | Proxied via the backend Node server with automatic reconnection and idle recycling. |
| Cloudflare Pages | Browser Direct | Static CDN | Browser connects directly to official Google endpoints using the locally configured API key. |
⚙️ Docker Environment Variables:
ENABLE_LIVE_WS_PROXY=true: Enables backend WebSocket proxy.RUNTIME_LIVE_API_BASE_URL=/api/live: Points client to proxy endpoint.LIVE_WS_IDLE_TIMEOUT_MS=300000: Automatically recycles connections idle for 5 minutes.