The announcement separates three jobs: talking, translating and turning ongoing speech into text.
What happened
GPT-Realtime-2 adds configurable reasoning to speech conversations. OpenAI also introduced GPT-Realtime-Translate and GPT-Realtime-Whisper through its Realtime API.
Why it matters
Developers can match a model to the interaction they need instead of treating every audio workflow as a chatbot. OpenAI positions the releases for applications that respond while a conversation is still unfolding.
The bigger picture
Our take. Realistic testing should include interruptions, background noise and names that matter to the task. A natural voice is only one part of a useful spoken interface.
Source checked November 18, 2026. Company claims are attributed; paragraphs marked “Our take” are editorial interpretation.