OpenAI WebRTC Audio Session, now with document context
Simon Willison revisited his OpenAI WebRTC audio playground, originally built in December 2024 to test the WebRTC API for realtime audio models. Last month, OpenAI introduced GPT-Realtime-2, promoted as "our first voice model with GPT‑5‑class reasoning" and a knowledge cut-off of September 30, 2024. Willison noted that this model has not yet appeared in the ChatGPT iPhone app, prompting him to update his browser-based tool. The updated playground allows users to select the new model and paste in document context for audio conversations about that information. The post is a link blog entry published on June 12, 2026, and is tagged with audio, tools, AI, OpenAI, generative AI, LLMs, multi-modal output, and WebRTC.
Developers can now experiment with GPT-Realtime-2's voice reasoning in a browser with custom document context.