OPENAI UPDATES CHATGPT VOICE TO ENABLE SIMULTANEOUS LISTENING AND SPEAKING

OpenAI has released updated voice technology for ChatGPT, introducing two new models called GPT-Live-1 and a mini version to all users. The rollout includes access across mobile applications and the website, with availability expanding over one to two days. According to Atty Eleti, product lead for ChatGPT voice, the primary improvement centres on more human-sounding audio with enhanced conversational qualities. Users can select from three different levels of intelligence to tailor response depth to their needs.

The core technical advancement is a framework termed continuous interaction, which permits the AI to receive information and generate output simultaneously. This represents a departure from previous voice modes that required users to finish speaking before the system could respond. Kundan Kumar, research lead for the GPT-Live model, stated that the architecture can delegate complex reasoning tasks to GPT-5.5 whilst maintaining active conversation with the user, mirroring human interaction patterns. The system can also generate visual answers such as weather reports or sports scores when Eleti noted that display format serves the response better than spoken output.

A significant practical application is live translation capability. Users speaking in English can receive real-time translation into Spanish, Hindi or other languages with minimal delay. The voice mode now acknowledges users as they speak and adjusts its pace if requested to do so. All ChatGPT users, including those on free and paid accounts, have access to these updated voice features.