gemini-live-api

Real-time audio and video chat with the Gemini Live API is enabled through bidirectional streaming over WebSocket connections. The Skill provides scripts to handle low-latency audio, video, and text conversations directly between client applications and Gemini models.

Voice activity detection (VAD) monitors audio input to identify when a user speaks. The Gemini Live API uses VAD for interruption handling, automatically pausing or stopping the model's audio output when a user interrupts the ongoing conversation.

Yes, the Gemini Live API supports function calling and tool integration during active WebSocket conversations. This allows the model to trigger external tools and retrieve real-time data while maintaining a persistent, low-latency streaming session.

Client-side connections are secured using ephemeral tokens provided by the Gemini Live API. These tokens enable secure WebSocket connections and support session resumption, ensuring persistent and safe interactions across devices like mobile apps and kiosks.

Yes, the Skill requires the google-genai and pyaudio dependencies to process real-time audio input and output. These libraries handle the bidirectional audio streaming and model communication necessary for live multimodal conversations.

Low-latency multimodal streaming with Gemini is applicable to customer support hotlines, hands-free assistants, and interactive virtual agents. It supports immersive experiences across mobile apps, chatbots, and kiosks requiring real-time voice, video, and text interactions.