Summary & Key Takeaways ​

• The OpenAI WebRTC audio tool has been updated by Simon Willison.
• It now supports the new GPT-Realtime-2 model, offering GPT-5-class reasoning.
• The model has a knowledge cut-off of September 30, 2024.
• A key new feature is the ability to paste document context for audio conversations.
• This allows users to discuss specific information with the AI in real-time.
• The GPT-Realtime-2 model is not yet available in the ChatGPT iPhone app.

Our Commentary ​

I've been waiting for this kind of integration! The idea of feeding an AI a document and then having a real-time audio conversation about it is incredibly powerful. It feels like a step closer to truly interactive knowledge exploration. Simon Willison's work here is a fantastic example of how community developers push the boundaries of what's possible with new AI APIs. We're excited to see what people build with this.