OpenAI has released GPT-Live-1 and GPT-Live-1 mini, two conversational models designed to handle interruptions and manage turn-taking in live speech. These are full-duplex systems, allowing users to interrupt the assistant naturally and enabling features such as real-time translation. The company is replacing the existing Advanced Voice Mode in ChatGPT with GPT-Live-1 mini as the default option. Paid subscribers will gain access to the larger GPT-Live-1 model. The previous system combined a speech-to-text engine, a large language model for generating answers, and a text-to-speech engine for delivery.
How the technology works
According to a press briefing, the new models address issues where assistants interrupt speakers or lack the intelligence to answer complex questions. OpenAI’s new systems send queries to its latest text models, such as GPT-5.5, for search, reasoning, or agentic capabilities while the conversation continues. The model can also remain silent for extended periods, absorbing context until required. Because the voice mode accesses newer GPT models, it can present information in visual formats. Other startups, including Monogram, which raised $40 million in seed funding from DST and Lux Capital, are also adopting visual responses to increase interactivity.
Longer conversations
OpenAI states the new voice mode in ChatGPT is designed for extended dialogue. Atty Eleti, product lead for ChatGPT Voice, noted during the briefing that he has held 30- to 40-minute-long conversations using the voice feature while walking. The company believes voice could become the primary interface for computing in complex tasks. Reports suggest the firm may launch a pair of earbuds with AI capabilities this year, though it provided no details on hardware products.
“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.
OpenAI has spent the past few years improving voice features to make ChatGPT’s voice mode sound more natural. The company reported that more than 150 million people speak to ChatGPT using features like Voice and Dictation. Rivals are also working to make assistants more expressive. Both Apple and Amazon have updated their assistants to improve conversational flow and context handling. Startups like Sesame, founded by Oculus co-founder Brendan Iribe and Ankit Kumar, have launched AI assistants that maintain natural conversation while completing background tasks.
Limitations and safety
OpenAI aims to let users talk to its assistant hands-free for longer durations. Despite claims that the new mode sounds more natural, the company emphasised it is not aiming to create an AI companion. The new models include safeguards to provide age-appropriate responses to teens and offer resources if the conversation turns to topics like self-harm. During the demo, the live translation feature in Hindi produced a heavy American accent and an unnatural, slightly bookish tone. The company stated the new mode is optimised for “most spoken languages” but did not specify which ones.
What it means
Users can expect shorter wait times and the ability to interrupt without the conversation resetting. The shift to visual outputs means assistants can now show data or diagrams alongside spoken answers. However, non-native speakers should expect some robotic quality in translated speech, particularly in languages other than English.
