OpenAI launches new voice models for more natural live conversations

OpenAI today released new conversation models, called GPT-Live-1 and GPT-Live-1 mini, claiming that they feel more natural and can handle turn-taking better. These are full-duplex models, meaning they can talk and listen at the same time, allowing users to interrupt naturally and enable features like live subtitles.

The company is also replacing ChatGPT’s current advanced audio mode with GPT-Live-1 mini by default. Users of paid tiers will have access to the larger GPT-Live-1 model. The former combines a speech-to-text model to transcribe speech, a large language model to generate responses, and a text-to-speech model to provide the final answer.

The company said in a press conference that the new models solve problems such as users being interrupted while they are speaking and not having enough intelligence to answer questions. The new OpenAI models will send the query to their state-of-the-art textual models such as GPT-5.5 for search, inference, or agent capabilities while continuing the conversation.

OpenAI also showed that the model can remain silent for an extended period and absorb the context of the conversation until called upon. In addition, since the new audio mode has access to newer GPT models, it can also present some information in a visual format. Other startups like Monogram, Which raised $40 million in seed funding from DST and Lux ​​CapitalThey also tend to visual responses to make assistants more interactive.

The company said that the new voice mode in ChatGPT is designed for longer conversations. During the press conference, Atty Eleti, product lead for ChatGPT Voice, said he has had 30-40 minute conversations with the voice feature while walking.

OpenAI believes that voice can be the primary interface to computing in complex businesses. Reports indicated the possibility of its launch A pair of earbuds with AI capabilities this year. However, it did not provide any information about the hardware products.

“Over time, we think this will also open up the ability to use voice as a kind of core interface to computing, managing long-running and increasingly complex proxy work. These are the kind of amazing use cases that we’re seeing people using Codex and ChatGPT to accomplish, and we think voice can be the future interface for all kinds of work,” Illetti said.

OpenAI has been working on support Audio based Features Over the past few years to make voice mode in ChatGPT seem more natural. The company said more than 150 million people talk to ChatGPT using features like voice and dictation.

Competitors are also trying to make assistants more expressive.

both of them apple and Amazon They updated their assistants to be more conversational with better context handling. Start-up companies such as sesamefounded by Oculus co-founders Brendan Iribe and Ankit Kumar, has also launched AI assistants with more natural conversation while completing tasks in the background.

OpenAI is moving in the same direction, aiming to let users talk to its assistant hands-free for longer. Although it claims that the new voice mode sounds more natural, the company has stressed that it does not aim to make this mode a companion to artificial intelligence. She noted that the new models have built-in safeguards to give age-appropriate responses to teens and provide resources if the conversation turns to topics like self-harm.

The new voice mode still needs work. During the demo, when the company demonstrated the live translation feature in Hindi, the assistant had a heavy American accent and spoke in Hindi that sounded unnatural and had a slightly biblical accent. The company said the new mode is optimized for “most spoken languages” but did not specify which ones.