OpenAI's GPT-Live brings voice-controlled multi-agent coordination to desktop

OpenAI has expanded ChatGPT’s desktop app with a major voice update for business and education users. The new ChatGPT Voice experience is now rolling out to Enterprise, EDU, and Business workspaces, offering two distinct modes. First, Voice in Chat enables natural, real-time conversations — powered by GPT-Live — with support for interruptions, follow-up questions, and web search. This works on desktop, web, iOS, and Android. Second, and more notably, Voice in Work and Codex lets users control their computer and coordinate multiple AI agents through a single conversation. On macOS or Windows, users can start tasks, check progress, interrupt or redirect work, and receive updates while agents run in the background.

Under the hood, OpenAI’s GPT-Live model handles the real-time voice interactions, making the experience feel fluid and responsive. During an initial two-week period, workspace owners must manually enable Advanced voice capabilities and Early Model Access for members to select the Live option. After that, GPT-Live becomes the default voice experience. For credit-based workspaces, Voice in Chat costs 5 credits per minute, while Voice in Work and Codex runs at approximately 6 credits per minute. Legacy Enterprise plans include bundled voice minutes per window. Tasks delegated through Work or Codex also consume credits from the shared usage pool at standard rates.

The move signals OpenAI’s push to make AI agents more accessible for complex, multi-step workflows. By allowing voice commands to orchestrate agents, check progress, and manage projects, professionals can keep work moving without constantly switching apps or typing commands. The addition of iOS remote access further extends utility, letting users monitor and redirect agents from their phone. This update positions ChatGPT as a central hub for voice-driven task coordination in enterprise settings.