OpenAI launches GPT-Live, its most natural voice AI for ChatGPT yet

OpenAI has introduced a new family of speech models that it says will make conversations with ChatGPT feel significantly closer to speaking with another person. The company on Wednesday announced GPT-Live-1 and GPT-Live-1 mini, replacing the existing Advanced Voice Mode with a system built to handle interruptions more naturally, respond with lower latency and support continuous, real-time dialogue.

The launch marks another step in OpenAI's broader effort to make voice a central way of interacting with artificial intelligence rather than a secondary feature. The company believes improvements in conversational AI will eventually allow users to complete increasingly sophisticated tasks through speech alone, while keeping access to more advanced reasoning capabilities in the background.

More from Tech

GPT-Live: A new approach to voice conversations

Unlike the earlier version of ChatGPT Voice, which relied on separate speech recognition, language generation and speech synthesis models, the new GPT-Live family is designed as a full-duplex system. That means it can process incoming speech while speaking at the same time, making exchanges feel less rigid and allowing users to interrupt naturally without breaking the flow of conversation.

OpenAI said GPT-Live-1 mini will become the default voice model inside ChatGPT, while subscribers on paid plans will gain access to the more capable GPT-Live-1.

The company said the redesigned architecture addresses some of the biggest complaints about AI voice assistants, including awkward interruptions, delayed responses and limited conversational intelligence. During a media briefing, OpenAI explained that the voice models can seamlessly call on its latest text models, including GPT-5.5, whenever search, reasoning or agentic capabilities are required, while keeping the spoken conversation uninterrupted.

Another capability demonstrated during the announcement was the model's ability to remain silent for extended periods while continuing to absorb conversational context before responding when prompted. OpenAI also said the upgraded voice experience can display visual information where appropriate, extending interactions beyond spoken replies.

The company believes users are already treating ChatGPT as a long-form conversational assistant rather than simply asking short questions.

Speaking during the briefing, ChatGPT Voice Product Lead Atty Eleti said he regularly spends between 30 and 40 minutes speaking with ChatGPT while out on walks, suggesting that users are becoming increasingly comfortable with extended voice sessions.

Looking ahead, OpenAI sees voice becoming a primary interface for computing.

"Over time, we think this will also unlock the ability to use voice as a kind of primary interface to computing, and to manage increasingly complex long-running agentic work. The kind of amazing use cases that we see people using Codex and ChatGPT to accomplish, we think voice can be the future interface to all kinds of work," Eleti said, according to the media reports.

The announcement also arrives amid reports that OpenAI is exploring dedicated AI hardware, including AI-enabled earbuds. While the company did not comment on future hardware products, its emphasis on hands-free interaction suggests voice is becoming an increasingly important part of its long-term strategy.

According to OpenAI, more than 150 million people already use ChatGPT's Voice and Dictation features.

Competition grows, but challenges remain

OpenAI is not alone in trying to build more conversational AI assistants. Technology companies including Apple and Amazon have recently upgraded their digital assistants with stronger contextual understanding and more natural dialogue. Meanwhile, startups are also experimenting with richer voice interfaces. Among them is Monogram, which recently secured $40 million in seed funding from DST and Lux Capital and is focusing on assistants that combine spoken responses with visual information. Another entrant, Sesame, founded by Oculus co-founder Brendan Iribe and Ankit Kumar, has also developed conversational AI designed to carry out tasks while maintaining natural interactions.

Even so, OpenAI acknowledged that its latest voice technology is still evolving. During a demonstration of live Hindi translation, the assistant spoke with a distinctly American accent and produced Hindi that sounded formal and somewhat unnatural. Although the company said the new models are optimised for the world's most widely spoken languages, it did not identify exactly which languages have received that level of optimisation.

OpenAI also stressed that its objective is to build a capable AI assistant rather than an AI companion. The company said safeguards have been built into the new voice experience to provide age-appropriate responses for teenagers and to direct users towards appropriate support resources if conversations involve subjects such as self-harm.

With GPT-Live now replacing the previous voice system inside ChatGPT, OpenAI is betting that conversational AI will become an increasingly common way for people to search for information, complete work and interact with intelligent software without needing a keyboard or screen for every task.