ChatGPT Live could make talking to AI feel straight out of the movies

AI voice assistants have been chasing the sci-fi dream for years, but they still have a hard time holding a conversation with humans. Most voice systems still need clear turns, clean pauses, and a few seconds before they respond. OpenAI is now rolling out GPT-Live, a new voice model for ChatGPT Voice that is designed to make those exchanges feel faster and less scripted.

The main upgrade is what OpenAI calls a full-duplex architecture. In simpler terms, GPT-Live can listen and speak at the same time. It continuously processes what the user is saying while also generating its own response, allowing it to decide when to talk, when to pause, when to keep listening, and when to use a tool.

What makes GPT-Live different from older voice systems?

OpenAI says GPT-Live can handle the messier parts of a normal conversation more smoothly. Users can interrupt ChatGPT with another question, pause while thinking, ask it to slow down, or tell it to stay quiet and listen. It can also respond with small acknowledgments like “mhmm” or “got it,” so the conversation does not feel as rigid.

The model is also designed to work better in noisy environments. The company says ChatGPT Voice should now be better at focusing on the user’s voice when there is background noise, such as traffic or nearby conversations.

GPT-Live can also hand off tougher requests in the background. If a question needs web search, deeper reasoning, or more complex work, it can delegate that task to GPT-5.5 and bring the result back when it is ready. The spoken conversation can continue while that happens.

Who gets GPT-Live first?

GPT-Live is rolling out globally in ChatGPT on iOS, Android, and the web. GPT-Live-1 will power ChatGPT Voice for Go, Plus, and Pro users. Free users will get GPT-Live-1 mini instead.

OpenAI is also adding visual cards to voice chats. So, if you ask about weather, stocks, sports, or similar topics, ChatGPT can show a card on screen while continuing the voice conversation. Search, memory, images, and file uploads will also continue to work with ChatGPT Voice.

Currently, the main limitation is around video. GPT-Live can handle voice conversations inside ChatGPT at launch, but it does not yet work with video or screen sharing. In other words, you cannot point your camera at something or share your screen during a GPT-Live voice chat yet. OpenAI says those features will arrive later.