OpenAI Brings GPT-Live Voice Mode to ChatGPT Desktop App for Mac

OpenAI is rolling out ChatGPT Voice to its desktop app today, restoring a full voice mode to the Mac app for the first time since the feature was removed from the older ChatGPT client. The update is powered by GPT-Live, the upgraded voice architecture that debuted on OpenAI's mobile apps earlier this month.

What's new

According to OpenAI's release notes, ChatGPT Voice lets users "talk through work and coordinate tasks in Chat, Work, and Codex" directly within the desktop app. Users can start a new chat or task in voice mode, then ask ChatGPT to start, check, or steer work running in other threads — effectively directing multiple concurrent agent sessions by speaking rather than typing.

On macOS specifically, a new "Screen context" toggle lets ChatGPT see an appshot of the user's frontmost window, giving the voice assistant visual context about what's currently on screen.

OpenAI confirmed the feature is available on Plus, Pro, Business, Edu, and Enterprise plans on desktop, rolling out globally starting today. The company also said the new voice mode works with ChatGPT's Remote feature on iOS, allowing cross-device continuity between phone and desktop sessions.

OpenAI engineer Guinness Chen, who said he had been testing the feature for several weeks, claimed on X that "anything you can do with Codex, you can now drive through voice," adding that GPT-Live makes the conversation "feel alive" while Codex provides the ability to act on commands. That framing has not been independently verified and reflects the company's internal testing experience rather than confirmed third-party benchmarks.

Multi-folder project support

Alongside voice, OpenAI shipped an update to how local projects work in the desktop app. Projects can now reference multiple related folders instead of a single directory. From a project's menu, users select "Edit project" to add folders and designate a primary folder. New chats, Git operations, and automatic discovery of AGENTS.md, skills, and config.toml files all use the primary folder, while secondary folders remain accessible for file search, reading, and editing.

Background

This update follows OpenAI's recent rollout of GPT-5.6 models and the ChatGPT Work agent, part of a broader push to reposition the Codex-based desktop client as the primary ChatGPT experience on Mac. Voice mode had been absent from the standalone Mac app for some time and never received the GPT-Live upgrade that iPhone and iPad users got earlier this year — until now.

Pricing for the underlying GPT-Live voice model has not been separately disclosed; access is bundled into existing ChatGPT subscription tiers.

What this means

This update signals OpenAI's continued push to make ChatGPT a hands-free control layer for agentic work rather than just a chat interface. By tying voice directly to Codex and Work agents, OpenAI is betting that developers and knowledge workers will want to supervise and steer multiple background tasks conversationally, similar to managing a team through spoken check-ins. The bigger open question is whether bundling Codex-style agent control into the consumer-facing Mac app — at the cost of native app feel and size, as OpenAI has acknowledged — will pay off in daily usage, or whether power users will find voice-driven agent management more novelty than necessity.