OpenAI today released new conversation models called GPT-Live-1 and GPT-Live-1 mini, which it claims sound more natural and handle turn-taking better. These are full duplex models. This means you can speak and listen at the same time, allowing users to pause naturally and enabling features like live translation.
The company will also replace ChatGPT’s current advanced audio mode with GPT-Live-1 mini by default. Paid tier users will have access to the larger GPT-Live-1 model. Previous models combined a text-to-speech synthesis model to transcribe the speech, a large-scale language model to generate the response, and a text-to-speech synthesis model to provide the final answer.
The company said in a press conference that the new model solves problems such as interrupting a user’s conversation or not having enough intelligence to answer a question. OpenAI’s new models continue the conversation while querying modern text models such as GPT-5.5 for search, inference, or agent functionality.
OpenAI also showed that the model can remain silent for long periods of time and absorb the context of the conversation until it is called. In addition, the new voice mode has access to the new GPT model, so you can also display some information in a visual format. Other startups like Monogram, which raised $40 million in seed funding from DST and Lux Capital, are also leaning into visual responses to make their assistants more interactive.
The company says ChatGPT’s new voice mode is designed to allow for longer conversations. Atty Eleti, head of product for ChatGPT Voice, said in a briefing that he has used the voice feature to converse for 30 to 40 minutes while walking.
OpenAI believes voice has the potential to become the primary interface to computing for complex tasks. According to reports, the company may release earphones equipped with AI features this year. However, no information about hardware products was provided.
“Over time, we think this will allow us to use voice as one of the primary interfaces to computing to manage increasingly complex and long-lasting agent work. Similar to the amazing use cases that people are accomplishing with Codex and ChatGPT, we think voice can be the future interface for all kinds of work,” said Eleti.
OpenAI has been working on voice-based feature enhancements over the past few years to make ChatGPT’s voice mode sound more natural. The company says more than 150 million people have conversations with ChatGPT using features such as voice and dictation.
Rival companies are also trying to improve the expressiveness of their assistants.
Apple and Amazon have both updated their Assistants to improve context handling and make them more conversational. Startups like Sesame, founded by Oculus co-founders Brendan Iribe and Ankit Kumar, have also launched AI assistants that conduct more natural conversations while completing tasks in the background.
OpenAI is moving in the same direction, aiming to allow users to have long, hands-free conversations with their assistants. Despite the company’s claims that the new voice mode sounds more natural, the company stressed that it’s not aiming to be an AI companion. He noted that the new model includes safeguards to provide age-appropriate treatment to teens and provide resources if the conversation turns to topics such as self-harm.
New voice modes still need work. During a demo, the company showed off its live translation feature in Hindi, and the assistant spoke in Hindi with a strong American accent, which sounded unnatural and a bit bookish. The company said the new mode is optimized for the “most commonly used languages,” but did not specify which ones.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
