OpenAI's GPT-4 Realtime model now lets you upload documents and discuss them by voice in a web browser.

What happened: A developer rebuilt their WebRTC audio playground tool to support OpenAI's new GPT-Realtime-2 model (described by OpenAI as their first voice model with GPT-5-class reasoning) and added the ability to paste in document context for audio conversations.

Why it matters: GPT-Realtime-2 offers a more capable voice interaction experience than the earlier WebRTC API model, with the added ability to ground conversations in specific documents—making voice a practical way to explore information interactively rather than just chat.

What to watch: The tool is available in a web browser now; however, the GPT-Realtime-2 model has not yet appeared in the ChatGPT iPhone app despite its announcement last month.