OpenAI Realtime API Now Commercially Available: Low-Latency Voice AI Interaction Opens Up Ecosystem

OpenAI's Realtime API is now open to all developers, supporting bidirectional voice conversations with latency under 200ms, built-in voice activity detection (VAD), and endpoint detection. Compared to traditional ASR+LLM+TTS pipelines, it better preserves emotion and tone. The GPT-4o Audio model supports voice input and output, priced at $0.10 per minute for input and $0.20 per minute for output. A wide range of voice AI applications will be built on this foundation.