OpenAI is putting GPT-Live voice control inside the ChatGPT desktop app, letting users direct chats, Work tasks and Codex agents by speaking instead of typing.
The update is rolling out July 23 for desktop users on macOS and Windows, according to 9to5Mac . The most immediate audience is people who already keep ChatGPT open while working: developers, analysts, students and enterprise users who want the assistant to act across active tasks without stopping for typed prompts.
OpenAI gives desktop users a voice layer for Chat, Work and Codex
ChatGPT Voice on desktop is powered by GPT-Live, OpenAI’s upgraded voice mode that arrived on mobile earlier this month. The release notes describe the feature as a way to “talk through work and coordinate tasks in Chat, Work, and Codex in the ChatGPT desktop app.”
That wording matters. This is not just voice dictation pasted into a chatbot box. OpenAI is positioning voice as a control layer for ongoing work inside the desktop app.
“Start a new chat or task in voice mode, then ask ChatGPT to start, check, or steer work in other threads.”
The practical question: can spoken instructions become a faster way to manage multiple AI tasks than switching between threads and typing status checks?
OpenAI’s own post, quoted by 9to5Mac, frames the update in more agentic terms:
“ChatGPT Voice is now in the desktop app. Control your computer and direct multiple agents running in ChatGPT Work or Codex, using just your voice. It's powered by GPT-Live, so it can speak, listen, and coordinate work in the app at the same time. Rolling out globally today…”
The desktop version also connects to several existing ChatGPT capabilities. The release notes say ChatGPT Voice works with OpenAI’s Computer Use feature, local files and ChatGPT plugins. It also works with ChatGPT’s Remote feature on iOS.
That combination gives the desktop rollout more weight than a standalone voice feature. Voice is being tied to the parts of ChatGPT that can inspect files, coordinate work and interact with computer tasks.
Builders get voice control over Codex, plus cleaner local project folders
For developers, the biggest change is the link between ChatGPT Voice and Codex. OpenAI says users can direct multiple agents running in ChatGPT Work or Codex by voice.
That turns the desktop app into a more active coding workspace. A user could start a task in voice mode, ask ChatGPT to check progress in another thread, or steer work that is already running. The source does not provide benchmarks, latency figures or examples of completed coding tasks, so the performance ceiling is still unproven.
The update also improves local project handling. Local projects in the ChatGPT desktop app can now include multiple related folders, according to the release notes.
From a project menu, users can select Edit project to add folders and choose the primary folder. New chats, Git operations and automatic discovery of AGENTS.md, skills and config.toml use that primary folder. Secondary folders remain available for file search, reading and editing.
| Desktop project feature | How the update changes it |
|---|---|
| Multiple folders | Local projects can include more than one related folder |
| Primary folder | New chats, Git operations and automatic discovery use it |
| Secondary folders | Available for file search, reading and editing |
| Voice control | Can start, check or steer work across threads |
The developer question: does voice become a useful layer for reviewing and steering agent work, or does serious coding still collapse back into typed prompts and terminal-level checks?
OpenAI’s timing also lands shortly after its Build Week materials described GPT-5.6 and Codex as available in ChatGPT, with the Devpost Hackathons plugin running inside ChatGPT desktop or mobile. That context points to the same product direction: keep builders inside ChatGPT while they plan, code and submit work.
Mac users get Appshots context, while Windows gets the core voice mode
Both macOS and Windows get ChatGPT Voice. Mac users, however, get one extra capability through Appshots.
When used with ChatGPT Voice, Appshots lets ChatGPT reference the active window in focus for better context. In plain terms, the assistant can use what is currently on screen as part of the conversation, rather than relying only on what the user says or pastes.
That is the clearest platform difference in the release notes.
| Feature | macOS | Windows |
|---|---|---|
| ChatGPT Voice | Yes | Yes |
| GPT-Live-powered conversations | Yes | Yes |
| Codex and Work coordination | Yes | Yes |
| Appshots active-window context | Yes | Not specified in the source |
The user question: will active-window context make voice feel less like a chatbot and more like a working assistant that understands the task already on screen?
Access is not universal across every ChatGPT account. The release notes say voice is supported on OpenAI plans including Plus, Pro, Business, Edu and Enterprise on desktop.
9to5Mac also notes that voice mode had been removed from the old Mac app and that the old app never received the newer GPT-Live version available on ChatGPT for iPhone and iPad. As of this update, the Mac app has a first-class voice mode again.
For related MLXIO coverage of AI features moving into everyday software, see Adobe Slips AI Editing Into Project Indigo Camera App. For a reminder that AI claims need clear evidence, see Key Trends Analysis Makes Big Claims With No Evidence.
The next test is reliability, not feature count
OpenAI has answered the availability question at a high level: the rollout is global today, and desktop support covers macOS and Windows for the listed plans. The release notes do not say whether every eligible user receives the feature at the same moment.
The harder questions are practical. Voice work depends on latency, interruption handling, transcription accuracy and whether the model can keep context while coordinating several threads. The source does not provide performance data on any of those points.
Enterprise use also remains a watch item. The update supports Business, Edu and Enterprise plans, but the release notes cited by 9to5Mac do not detail new admin controls, company data settings or deployment policies specific to desktop voice.
The signal is clear enough, though: OpenAI wants the ChatGPT desktop app to become a place where users direct ongoing work, not just ask isolated questions. If GPT-Live can reliably steer Codex, local files and active app context by voice, the desktop app becomes harder to treat as a side window — and easier to treat as a daily work surface.
What This Means For You
- Desktop users on macOS and Windows can control ChatGPT workflows by voice starting with the July 23 rollout.
- The update turns voice into a task-control layer for Chat, Work, Codex, Computer Use, local files and plugins.
- Developers, analysts, students and enterprise users may manage multiple AI tasks faster without switching threads or typing prompts.
