OpenAI Realtime API

// about OpenAI Realtime API

OpenAI's Realtime API enables direct, low-latency speech-to-speech conversations with GPT-4o, without the latency of a traditional transcription → LLM → TTS pipeline. Applications built on it can respond to voice in under 300ms, support natural interruptions mid-sentence, and detect tone and emotion in the speaker's voice. It enables a new class of voice AI products — AI phone agents, real-time translators, voice-controlled assistants — that feel genuinely conversational rather than like talking to an automated phone menu.