GPT-Live: OpenAI's Human-Like Voice AI for ChatGPT

OpenAI has launched GPT-Live, a new set of voice AI models for ChatGPT. The key difference is that GPT-Live can listen and respond at the same time, unlike older systems that waited for users to finish speaking. This rollout started globally on Wednesday, with paid subscribers getting access first. OpenAI announced the launch on X, calling it a "new generation of voice models for natural human-AI interaction." It's available to ChatGPT users on Go, Plus, and Pro plans, requiring the latest app version on iOS and Android for mobile users. Paid users get GPT-Live-1, while free users receive GPT-Live-1 mini. What's interesting is that while GPT-Live handles live conversations, it can pass more complex tasks to a background AI model, specifically GPT-5.5 at launch. This means web searches and deeper reasoning are handled by a more powerful system. OpenAI also states that users can now interrupt the AI, ask it to slow down, or simply listen without an immediate response. The system is also better at filtering out background noise. ChatGPT's nine voice options have been updated, and for certain topics like weather or sports, visual cards can now be displayed while users continue speaking. This feature also supports search, memory, and file uploads. This development could significantly change how people interact with AI in daily conversations.