OpenAI updated ChatGPT Voice Mode — the changes

OpenAI has announced an update to the Advanced Voice feature, the voice mode in ChatGPT that provides real-time communication with AI. The company aims to make the voice assistant more friendly in conversation and interrupt users less during pauses.

TechCrunch reported on it.

ChatGPT voice assistant will interrupt users less often

Manuka Strata, OpenAI's post-training researcher, shared details about the update. From now on, free ChatGPT users can pause their speech without the risk of being interrupted by the system. For ChatGPT Plus, Teams, Edu, Business, and Pro subscribers, the Advanced Voice mode has also become less intrusive and received an improved assistant personality.

According to an OpenAI representative, the new voice assistant for those who subscribe to paid tariff plans has become more direct, engaged in the conversation, concise, specific, and creative in its responses. These changes are intended to make communication with AI even more comfortable and useful.

At the same time, OpenAI is forced to act quickly as strong competitors are emerging in the market. Sesame, an Andreessen Horowitz-backed startup created by Oculus co-founder Brendan Iribe, has become known for its natural voice assistants Maya and Miles. Moreover, large corporations are also striving to take the lead: Amazon is preparing to introduce an updated version of Alexa based on its own large language model.

Thanks to the updated Advanced Voice Mode, users get an assistant that does not interrupt during a conversation and demonstrates a more mature communication style. Such changes are expected to help OpenAI maintain its leadership position in the growing voice assistant segment.

As a reminder, OpenAI has released the powerful o1-pro AI model in the API for developers. Despite the model's performance, it turned out to be the most expensive. For a million tokens, you need to pay from USD 150.

We also wrote that OpenAI has admitted that the GPT-4.5 model often "hallucinates". It invents facts, disguising them as real, in more than a third of cases.