OpenAI Real Time

Low-latency, multimodal API enabling real-time, bidirectional speech-to-speech and text interactions with models like GPT-4o.

The OpenAI Realtime API is a low-latency, session-based interface for building highly responsive, multimodal AI applications. It leverages WebRTC, WebSocket, and SIP connections to facilitate instant, bidirectional communication with models (e.g., `gpt-realtime`). The API natively supports speech-to-speech interactions, audio streaming, and advanced features like server-side Voice Activity Detection (VAD) and function calling. This makes it the core tool for developing voice agents, real-time translation services, and dynamic collaborative platforms.

https://platform.openai.com/docs/api-reference/realtime