Part 7/12:

The presenters shared a live demo where users could:

• Talk to Gemini and get natural, multimodal responses.

• Share screens, video, or audio inputs.

• Control interactions with voice commands, including interruptions.

The demo showcased the streaming capabilities—how voice chunks are processed asynchronously for near-instantaneous response, handling interruptions smoothly.

Key points highlighted:

• The client-side code, written in HTML and JavaScript, uses WebSocket connections to interact with Gemini's API.

• It manages multiple modalities: speech, video, screen sharing.

• Interruption handling is managed via VAD, enabling users to stop responses or switch tasks dynamically.

The entire setup demonstrated how easily developers can prototype multimodal experiences.