OpenAI updates ChatGPT desktop with GPT Voice: what it means for AI builders and operators

OpenAI has released a major update to the ChatGPT desktop app that brings GPT-Live-powered voice to macOS and Windows, letting users talk through work and coordinate tasks across Chat, Work, and Codex. For AI builders and operators, this shifts voice from a simple Q&A interface to a primary control layer for agent orchestration and multi-threaded workflows.

What happened

The ChatGPT desktop app now includes GPT Voice mode, powered by GPT-Live, OpenAI's upgraded voice model that arrived on mobile earlier this month. Users can start a new chat or task in voice mode, then ask ChatGPT to start, check, or steer work in other threads. The feature works with Computer Use, local files, and ChatGPT plugins.

On Mac, Appshots lets ChatGPT reference the active window in focus for better context when using voice. The update also improves local project management: projects can now include multiple related folders, with a primary folder designation for new chats, Git operations, and automatic discovery of AGENTS.md, skills, and config.toml.

Why AI builders should care

This update foregrounds voice as a primary interaction mode for complex, multi-threaded tasks and agent orchestration. Instead of typing prompts to kick off a code review, trigger a pull request, or coordinate multiple agents in ChatGPT Work, developers can now drive those actions through voice.

For teams building automation workflows, this changes the ergonomics of agent interaction. Voice can start, guide, and coordinate tasks across multiple threads and agents from the desktop app. The integration with Codex means anything you can do with Codex, you can now drive through voice, including reviewing codebases and managing background tasks.

Practical implications

Voice coordination across threads is the headline feature. You can start a task in one thread, check progress in another, and steer a third, all hands-free. The Appshots capability on Mac adds contextual awareness by referencing the active window, which matters for computer-use agents and interface analysis.

The multi-folder project update is also significant for developers. Projects can now include multiple related folders, with a primary folder for new chats, Git operations, and auto-discovery of AGENTS.md, skills, and config.toml. Secondary folders remain available for file search, reading, and editing.

Caveats

Evidence for this update comes primarily from release notes and press coverage. Features may vary by platform, plan, and local permissions. The local file access and AGENTS.md/config.toml discovery depend on project setup and may not work identically across all environments. Voice mode is not available on the free tier, and some users have noted the desktop app still lacks a native Mac feel despite the feature additions.