Talking to ChatGPT may soon feel less like using a chatbot and more like speaking to another person.
OpenAI has unveiled GPT-Live, a new family of real-time voice AI models that powers ChatGPT's voice experience. Unlike earlier voice assistants that waited for users to finish speaking before replying, GPT-Live can listen and speak at the same time, making interactions smoother and more conversational.
Announcing the launch, OpenAI said, "We're launching GPT-Live, a new generation of voice models that make talking with AI feel much more like having a real conversation."
The key innovation lies in how the technology works. GPT-Live is built on a full-duplex architecture, allowing it to understand pauses, interruptions and changes in the flow of a conversation.
ALSO READ | GPT5.6 To Be Out Soon: OpenAI's Most Capable AI Model Set For Rollout After US Govt Nod
Instead of cutting users off or waiting in silence, it can respond with simple acknowledgements such as "mhmm" or "yeah," continue listening or jump into quick back-and-forth exchanges when appropriate. The result is a voice experience that feels far more natural than previous AI interactions.
OpenAI says GPT-Live is its most capable voice model to date. When users ask questions that require web searches or deeper reasoning, the voice model hands the request to the company's latest frontier model running in the background while continuing the conversation.
At launch, GPT-Live routes these more advanced tasks to GPT-5.5, allowing users to keep talking instead of waiting for the AI to process information.
The company has started rolling out two versions of the new model globally. GPT-Live-1 is available for paid ChatGPT users, while GPT-Live-1 mini powers the experience for users on the free tier.
The following features are being introduced across ChatGPT on the web, Android and iOS, with API support expected in the future.
Beyond answering questions, GPT-Live can perform live translations, work with text and images during conversations, access memory and web search when needed, and display spoken responses alongside streamed text.
These additions are designed to let users switch naturally between speaking, reading and viewing information without breaking the flow of the interaction.
ALSO READ | Meta To Launch Custom AI Chip 'Iris' In September To Cut Nvidia, AMD Reliance
OpenAI has also introduced new safety measures alongside GPT-Live. The company says the model has been trained with enhanced safeguards for sensitive conversations, age-appropriate behaviour for younger users and expert-reviewed responses for crisis-related situations.
The new launch reflects the company's broader vision of making voice the most intuitive way to interact with artificial intelligence. Rather than treating voice as an add-on, the company sees GPT-Live as a step towards AI that can hold continuous conversations while carrying out increasingly complex tasks behind the scenes.
"Our vision is to enable truly natural human-AI interaction: a world where collaborating with AI feels as fluid and responsive as working with another person, while reasoning and complex task execution happen seamlessly in the background," OpenAI said.
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.