An article summarized by TechCrunch:
OpenAI has introduced two new conversational AI voice models, GPT-Live-1 and GPT-Live-1 mini, designed to make conversations with ChatGPT feel much more natural. Unlike the previous voice system, the new models can listen and speak at the same time, allowing users to interrupt seamlessly, pause naturally, and even use features like live translation. GPT-Live-1 mini is now the default voice model in ChatGPT, while paid subscribers can access the more advanced GPT-Live-1.
The company says the upgraded voice mode fixes many of the frustrations users experienced with the previous version, such as cutting people off or struggling with more complex questions. The new system works alongside OpenAI's latest language models, including GPT-5.5, enabling stronger reasoning, web search, visual responses, and longer conversations. OpenAI believes voice could eventually become the primary way people interact with computers for everything from everyday questions to managing complex AI-powered tasks, noting that more than 150 million people already use ChatGPT's voice features.
Despite the improvements, OpenAI acknowledged the technology is still evolving. During a demonstration, its live Hindi translation sounded unnatural and carried a noticeable American accent, highlighting that multilingual voice capabilities still need refinement. While competitors like Apple, Amazon, and AI startups are also racing to create more conversational assistants, OpenAI emphasized that its goal is to build a useful voice interface, not an AI companion, and has added safeguards to provide age-appropriate responses and support for sensitive topics.
