A desktop update lets you speak commands to coordinate multiple AI agents and control apps. It hints at smoother voice driven workflows and faster multi agent coordination.
OpenAI unveiled an update to the ChatGPT desktop app on Thursday, adding the ChatGPT Voice voice mode to control AI agents on your computer.
The new capability is based on the ChatGPT-Live family of voice models launched by OpenAI earlier this month.
ChatGPT Voice works with ChatGPT Work and Codex, and notably enables using the computer’s capabilities to search for websites and applications. On macOS, via Appshots, users can grant the app access to what is displayed on the screen, including alt text.
At launch, the mobile version promised smoother conversations and better interruption handling, but it wasn’t intended to perform actions on mobile devices themselves. The desktop update broadens capabilities, enabling the dictation of complex command sequences and waiting for user input.
ChatGPT Voice is now in the desktop app. Control your computer and manage multiple agents working in ChatGPT Work or Codex using only your voice. It runs on GPT-Live and can speak, listen, and coordinate work within the app.
– OpenAI
Enhanced voice interaction on the desktop platform
In the demonstration video, the developer asked ChatGPT to create a new branch, make a pull request, and locate the root cause of an error in a single command.
Additionally, users can use ChatGPT Voice in Codex via remote access from the iOS app.
Anthropic has also updated Claude’s voice mode, capable of using the Opus, Sonnet, and Haiku models to carry out tasks in Gmail, Calendar, Slack, Notion, and Canva.
The feature is now rolling out globally worldwide, the company notes, highlighting the broad availability of the new capabilities. The update emphasizes the developers’ focus on computer control and coordinating the work of multiple agents using voice.
These steps illustrate the developers’ push to make voice-controlled computing more convenient and efficient, especially for coordinating the actions of multiple agents in real time.