OpenAI updates ChatGPT Voice with GPT-Live-1 model

In short: OpenAI is rolling out a new model for ChatGPT Voice called GPT-Live-1 that is designed to feel more like a real conversation and interrupt people less.

What happened

OpenAI said it is overhauling ChatGPT’s voice mode with a new model named GPT-Live-1. The company says the new version will wait longer before speaking, and it will keep listening if you pause mid-sentence.

OpenAI also says GPT-Live-1 can talk and listen at the same time. This is sometimes called “full duplex” (like a phone call, where both people can speak and hear each other at once). In earlier versions, ChatGPT Voice worked more like taking turns, which could lead to awkward timing and interruptions.

When GPT-Live-1 needs to do harder tasks, OpenAI says it can hand off parts of your request to the company’s best text models, such as GPT-5.5. In simple terms, it is like asking a quiet researcher to look something up, and then having the voice assistant explain the answer out loud.

The update also adds features like real-time translation, so it can translate as you speak instead of waiting for you to stop. OpenAI says ChatGPT Voice can also be told to stay quiet until you call on it. For topics like weather, stocks, and sports, it may show AI-made visuals, such as a forecast or a score.

OpenAI says it added safeguards to avoid harmful replies and to end chats in higher-risk situations. It also says the system can offer crisis helpline support for self-harm topics and give age-appropriate responses for teens.

Why it matters

For everyday users, this update could make voice chats feel less frustrating, especially for translation and back-and-forth conversations. It also signals that OpenAI is putting more attention on safety as more people use chatbots for personal and sensitive topics.

Source: The Verge AI