On Thursday, OpenAI released an update to the ChatGPT desktop app that brings ChatGPT Voice to the computer. With the feature, users can start talking to the assistant to trigger AI agents and carry out tasks directly on the machine.
The new feature runs on ChatGPT-Live, the family of voice models introduced by the company earlier this month. It is the same foundation that had been used in the mobile version, now adapted for an environment with more freedom of action.
According to the company, ChatGPT Voice works together with ChatGPT Work and Codex, in addition to drawing on computer-use capabilities to navigate websites and apps. On macOS, Appshots allows the program to read the content displayed on the screen, including alt text.
The difference compared to the previous launch is its scope. In its smartphone debut, the focus was on more natural conversations and better handling of interruptions, without the assistant carrying out actions on the device.
The desktop version, in turn, was designed for chained tasks. The user dictates instructions with several steps, and the assistant responds with questions when it needs some definition to move forward.
OpenAI also said that ChatGPT Voice can be activated in Codex from the iOS app through remote access. In practice, the cellphone becomes a controller for work running on the computer.
The move is not happening in isolation. Anthropic has also revamped Claude’s voice mode, which now triggers the Opus, Sonnet, and Haiku models to complete activities in tools such as Gmail, Calendar, Slack, Notion, and Canva.
The race for assistants capable of taking action, and not just responding, gains a new chapter with these nearly simultaneous updates.
The race for assistants capable of acting, and not just responding, gains a new chapter with these almost simultaneous updates.
