OpenAI has begun releasing two updated voice models for ChatGPT that allow simultaneous listening and speaking. The changes aim to create more natural conversations.
OpenAI started rolling out GPT-Live-1 and GPT-Live-1 mini on July 8. The models use a duplex architecture that processes user input while generating responses at the same time.
Users can interrupt the AI at any point and ask it to slow down. The system also provides verbal acknowledgements such as "mhmm" and "got it" while listening.
The models support live translation and can delegate complex tasks to other systems including GPT-5.5. They generate visual cards for topics like weather and sports scores.
New safeguards address unsafe content and allow parents to disable voice mode for teens. Audio from conversations is stored for 30 days.
Access begins with paid subscribers on Android, iOS and the web, followed by free users.
