OpenAI announced on Thursday that it has enhanced its ChatGPT desktop application with voice support, enabling users to speak commands that control AI agents and perform various tasks on their computers. This feature leverages the company's newly introduced voice model family, ChatGPT-Live, which debuted earlier this month. According to OpenAI, the voice functionality is compatible with both ChatGPT Work and Codex, and it can also utilize computer-use capabilities to browse websites and applications. Additionally, on macOS, the app can access on-screen content—including alt-text—through a feature called Appshots.

When ChatGPT Voice first launched on smartphones, it offered smoother conversations with improved interruption handling, but it was not designed to execute actions on mobile devices. Thursday's desktop update is more robust, allowing users to dictate complex, multi-step commands and respond when the AI requires their input. In a demonstration video, OpenAI showed a developer instructing ChatGPT to create a new thread, make a pull request, and identify the root cause of a bug using a single voice command.

The company noted that users can access ChatGPT Voice within Codex from the iOS app via remote control. Similarly, Anthropic has updated its voice mode for Claude, which can now use the company's Opus, Sonnet, and Haiku models to complete tasks across applications like Gmail, Calendar, Slack, Notion, and Canva.