OpenAI announced on Thursday local time that it had updated the ChatGPT desktop app with support for ChatGPT Voice. Users can now converse directly with ChatGPT by voice, control the AI agent, and have it perform various tasks on their computers.
The new feature is based on OpenAI’s new ChatGPT-Live family of voice models. OpenAI launched this model family earlier this month.
OpenAI said that ChatGPT Voice supports both ChatGPT Work and Codex, and can also access computer-control capabilities to help users access websites and applications. On macOS, with Appshots, users can also allow ChatGPT to access content on the screen, including alt text.
When ChatGPT Voice first launched on mobile, it primarily focused on providing a smoother voice-conversation experience and improving how it handled users interrupting the AI, but it could not directly perform actions on phones at the time.
This desktop update offers stronger execution capabilities. Users can give ChatGPT complex, multi-step instructions by voice and respond promptly when ChatGPT needs them to provide information or confirm an action.
In a demonstration video, OpenAI showed a developer using a single instruction to have ChatGPT create a new code thread, submit a Pull Request, and identify the root cause of a bug.
OpenAI also said that users can use ChatGPT Voice in Codex through the remote-access feature of the iOS app.
Anthropic has also updated Claude’s voice mode. The feature can access Anthropic’s Opus, Sonnet, and Haiku models to complete various tasks in applications such as Gmail, Calendar, Slack, Notion, and Canva.