Vobiz
Console

Build an AI Voice Agent

Welcome to the definitive guide on building an AI Voice Agent using Vobiz. By the end of this guide, you will understand how to connect leading conversational AI platforms directly to real phone numbers.

Introduction

An AI Voice Agent is a conversational system that can listen, process speech, and reply with human-like audio over a telephone call. By pairing Vobiz’s robust SIP and telecom infrastructure with providers like Vapi, Retell AI, or ElevenLabs, you can build powerful inbound customer support systems or outbound calling campaigns.

Instead of manually handling WebRTC streams, Vobiz allows you to easily route telephone audio (RTP streams) to these AI providers out of the box using standard SIP or WebSocket connections.

How it Works

The architecture of an AI Voice Agent generally consists of three parts:

  1. The Telecom Layer (Vobiz): Provides the actual phone number (DID), handles the SIP trunking, and manages the lifecycle of the phone call.
  2. The Interconnect: Vobiz forwards the call to your server or directly to an AI integration platform using a SIP URI or a Stream (WebSocket).
  3. The AI Layer: The AI provider (like Vapi, Retell, Pipecat) runs Speech-to-Text (STT), routes context to an LLM (like OpenAI GPT-4), generates text, and converts it back using Text-to-Speech (TTS).

Prerequisites

Before setting up your first agent, ensure you have:

  • A registered Vobiz account.
  • An active phone number purchased in the Vobiz dashboard.
  • An account with your preferred AI Voice provider.