OpenAI has brought ChatGPT Voice to the ChatGPT desktop app, moving the feature beyond conversation and into computer-based task control. The update lets users talk to the app, direct AI agents, and handle work across ChatGPT Work and Codex.
What changed in the desktop app
OpenAI said on Thursday that the ChatGPT desktop app now supports ChatGPT Voice. The feature lets users speak to the app to control AI agents and carry out tasks on their computer.
The update is powered by OpenAI's new family of voice models called ChatGPT-Live, which launched earlier this month. In practical terms, that means the desktop app can listen, speak, and coordinate work inside the app at the same time.
This is a more action-oriented version of ChatGPT Voice than the smartphone launch described in the source. The smartphone version offered smoother conversations and better interruption handling, but it was not built to take action on smartphones.
Why desktop voice matters
The desktop environment gives ChatGPT Voice a broader role. Instead of only responding in a voice conversation, the feature can now be used to issue complex commands with multiple steps.
OpenAI said users can dictate those commands and respond when ChatGPT needs input. That matters because many computer tasks are not single-step requests. They often require context, follow-up decisions, and coordination across work already happening in the app.
The update supports both ChatGPT Work and Codex. It can also tap computer use skills to look up websites and apps, giving voice commands a path into tasks that depend on information outside a single chat window.
- ChatGPT Work: users can direct agents by voice inside the work-focused environment.
- Codex: users can control coding-related agent activity with spoken instructions.
- Computer use skills: the feature can look up websites and apps when a task calls for it.
- macOS Appshots: users can allow access to what is on their screen, including alt-text.
A developer-focused example
OpenAI showed the feature in a demo video with a developer using one spoken command to ask ChatGPT to create a new thread, make a pull request, and find the root cause for a bug.
That example shows the shift from voice as an interface for chatting to voice as an interface for directing work. The user does not have to break the request into separate typed prompts before the agent starts coordinating the steps.
The important detail is not just that the command is spoken. It is that the command includes several connected actions. The desktop app can then respond when it needs more input from the user, keeping the interaction moving without reducing the workflow to a simple voice transcript.
How iOS fits into the rollout
OpenAI also said users can use ChatGPT Voice in Codex from the iOS app through remote access. The source does not describe the full remote access workflow, but it does show that the desktop update is not isolated from mobile use.
The distinction remains important: the earlier smartphone version of ChatGPT Voice was centered on conversation quality, including smoother exchanges and better interruption handling. The new desktop update is described as more capable because it can take action on computer-based tasks.
For users who already work with ChatGPT Work or Codex, the change makes voice another control layer for agent activity. For OpenAI, it also turns the desktop app into a more central place for coordinating spoken instructions, agent work, and computer context.
The broader voice assistant race
OpenAI is not the only company moving voice modes toward task completion. Anthropic has also updated its voice mode for Claude.
According to the source, Claude's voice mode can use Anthropic's Opus, Sonnet, and Haiku models to complete tasks in apps including Gmail, Calendar, Slack, Notion, and Canva.
That comparison points to a larger direction in AI products: voice is becoming less about asking a question aloud and more about managing work across tools. In OpenAI's case, the ChatGPT desktop app update ties that idea to ChatGPT Work, Codex, computer use skills, and macOS Appshots.
The result is a version of ChatGPT Voice that is designed for action on the desktop. Users can speak, the app can coordinate agents, and ChatGPT can ask for input when a task needs clarification.