GPT-4o mini Realtime Preview

Technical Specifications of gpt-4o-mini-realtime-preview

What is gpt-4o-mini-realtime-preview?

gpt-4o-mini-realtime-preview is a real-time multimodal model designed for fast, interactive experiences where users speak, type, or share visual input and expect immediate responses. It is well suited for applications that need live back-and-forth communication rather than standard single-turn request/response workflows.

The model can process speech, text, and images within the same experience, making it useful for assistants that listen to a caller, inspect on-screen or camera content, and respond in natural language or audio. Because it supports streaming input and output, developers can build systems that feel responsive during ongoing interactions instead of waiting for a full completion.

It also supports tool or function calling, which allows the model to trigger structured actions such as looking up data, calling backend services, or executing workflow steps. This makes gpt-4o-mini-realtime-preview a strong choice for grounded, action-oriented agents in customer support, operations, productivity, and multimodal assistant scenarios.

Main features of gpt-4o-mini-realtime-preview

• Real-time multimodal interaction: Accepts and responds across speech, text, and images for fluid live experiences.
• Bidirectional audio: Supports conversational voice interfaces where audio can be streamed in and responses can be streamed back out.
• Streaming responses: Delivers partial outputs incrementally, reducing perceived latency and improving responsiveness.
• Vision understanding: Interprets visual inputs such as camera frames, screenshots, or other images during a live session.
• Function and tool calling: Produces structured calls that let your application connect the model to business logic, databases, or external tools.
• Interactive agent behavior: Works well for assistants that must maintain turn-by-turn context during active sessions.
• Live call handling: Useful for phone or web-call scenarios involving fast speech understanding and immediate replies.
• Real-time captioning and transcription workflows: Can support experiences that convert ongoing speech into usable text in near real time.
• Structured outputs for actions: Helps applications turn conversational intent into reliable machine-readable instructions.
• Low-latency user experiences: Optimized for scenarios where responsiveness matters, such as support, coaching, monitoring, and guided workflows.

How to access and integrate gpt-4o-mini-realtime-preview

Step 1: Sign Up for API Key

First, create an account on CometAPI and generate your API key from the dashboard. This key is required to authenticate every request. Store it securely and avoid exposing it in client-side code or public repositories.

Step 2: Connect to gpt-4o-mini-realtime-preview API

The Realtime API uses WebSocket connections. Connect to CometAPI's WebSocket endpoint:

const ws = new WebSocket(
  "wss://api.cometapi.com/v1/realtime?model=gpt-4o-mini-realtime-preview",
  {
    headers: {
      "Authorization": "Bearer " + process.env.COMETAPI_API_KEY,
      "OpenAI-Beta": "realtime=v1"
    }
  }
);

ws.on("open", () => {
  ws.send(JSON.stringify({
    type: "session.update",
    session: {
      modalities: ["text", "audio"],
      instructions: "You are a helpful assistant."
    }
  }));
});

ws.on("message", (data) => {
  console.log(JSON.parse(data));
});

Step 3: Retrieve and Verify Results

The Realtime API streams responses through the WebSocket connection as server-sent events. Listen for response.audio.delta events for audio output and response.text.delta for text. Verify the session is established and responses are streaming correctly.