OpenAI brings voice control to coding tools on desktop

OpenAI has integrated its GPT-Live voice model with full-duplex capabilities into the ChatGPT desktop application for macOS and Windows. This integration allows users to control tools like Codex and ChatGPT Work using natural voice commands, eliminating the need for manual input and enabling hands-free management of complex coding tasks.

The technology is based on a continuous audio model capable of listening and speaking simultaneously, creating a more fluid conversation and eliminating rigid turn-taking. The model can delegate complex reasoning tasks to background models, such as GPT-5.5, while maintaining a natural conversational flow with the user. The new features allow software developers to orchestrate multiple concurrent coding jobs, review code requests, and debug applications using only voice commands.

For macOS, the desktop application incorporates "Appshots" and screen context features, allowing ChatGPT Voice to analyze the active window along with local files, codebase structures, and active plugins. This architecture creates a pair-programming dynamic where developers can discuss problems conversationally while agents perform tasks asynchronously. Developers can verbally convert design mockups into working code, splitting tasks across frontend, backend, and testing layers.

The service is available to paid subscribers on OpenAI's Plus, Pro, Business, Enterprise, and Education plans. While the system's comprehensive voice control offers significant benefits, its internal models and architectures remain proprietary and cannot be modified or self-hosted by users. Tasks initiated via voice commands directly consume existing quotas for Codex and ChatGPT Work.

The developer community has reacted with enthusiasm to the new features, highlighting the potential for hands-free orchestration of complex agent tasks, particularly during remote work or when not directly at the workstation.