OpenAI logo

AI Models

OpenAI Realtime

OpenAI's full-duplex Realtime speech API for live phone agents. Stream call audio in and play spoken replies back in the same session.

Speech-to-speechRealtimeTool callingWebSocketFull-duplex
Documentation
Where it fits

OpenAI Realtime in your voice stack.

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

What it contributes to the stack.

OpenAI Realtime is a full-duplex speech API for live agents. It keeps a persistent session so the model can listen, speak, interrupt, and call tools without a separate STT/TTS pipeline.

With AgentDuet, the phone or WhatsApp connection stays on AgentDuet. Your application bridges call audio into Realtime and plays spoken replies back to the caller.

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

Streams caller audio in and returns spoken model output over an open session.

Interruption handling

Supports barge-in so queued AgentDuet playback can be cleared immediately.

Tool calling

Requests approved application functions without leaving the live conversation.

Session control

Configures voice, turn detection, and modalities for the active realtime session.

Common use cases

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

Voice agents

Run continuous spoken conversations with low-latency replies on phone or WhatsApp.

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.

Related integrations

Compare nearby choices.

These providers occupy a similar role in the stack. Compare their strengths, operating model, and surrounding services.

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
Explore integration

Amazon Nova 2 Sonic

Amazon's speech-to-speech model on Bedrock for live phone agents. Stream call audio in and play spoken replies back in the same session.

Speech-to-speechAWSBedrockBarge-inTool use
Explore integration

Grok Voice

xAI's full-duplex Grok Voice model for live phone agents. Stream call audio in and play spoken replies back in the same session.

Speech-to-speechRealtimeTool callingWebSocketFull-duplex
Explore integration
Browse every integration

Build a voice agent.

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

Explore the SDK