Gemini logo

AI Models

Gemini Live

Google's real-time speech model for live phone agents. Stream call audio in and play spoken replies back in the same session.

Speech-to-speechMultimodalInterruptionTool callingNative audio
Documentation
Where it fits

Gemini Live in your voice stack.

  1. TelephonyPhone networkInbound call
  2. AgentDuetCall + audioAnswer, stream
  3. Speech-to-speechGemini LiveListens, call functions, replies
  4. ToolsActionsCRM, calendar, APIs
Overview

What it contributes to the stack.

Gemini Live is Google's real-time speech model built for live conversations. It listens and speaks in the same session, supports interruptions, and can request tools without dropping the call. That makes it a strong fit when you want a phone agent that sounds continuous, not turn-based IVR.

On AgentDuet, your application bridges live call audio to a Gemini Live session and plays speech back to the caller while AgentDuet handles answer and hangup.

For setup, credentials, and API details, use the Documentation link above.

Capabilities

The technical characteristics that matter when deciding whether this provider belongs in your application.

Bidirectional audio

Receives caller audio and emits generated speech concurrently over a live session.

Interruption handling

Reports interrupted responses so queued AgentDuet playback can be cleared.

Tool calling

Requests approved application functions without leaving the conversation.

Multimodal context

Can combine audio with text and supported visual input supplied by the application.

Common use cases

Practical scenarios where the provider's role is clear and the surrounding systems remain under application control.

Appointment coordination

Collect requirements, check availability through tools, and confirm a next step by voice.

Customer support

Handle routine questions while routing account actions through controlled systems.

Lead qualification

Conduct an adaptive conversation and capture structured outcomes for follow-up.

Build a voice agent.

AgentDuet handles the phone and messaging boundary. Your model runs the conversation.

Explore the SDK