At 1 AM today, OpenAI unveiled two groundbreaking advancements in its AI speech capabilities. Firstly, the introduction of the Realtime API, designed to facilitate the development of production-ready, real-time agents capable of seamless, low-latency multimodal interactions. Secondly, the unveiling of GPT-realtime, a cutting-edge speech-to-speech model that generates speech that is natural, fluent, and remarkably human-like, accurately mimicking human intonation, emotion, and speaking pace. This model is poised to revolutionize the creation of speech agents across various domains. Additionally, GPT-realtime introduces two novel, distinctive voices and significantly enhances its existing voice offerings.