Breaking

OpenAI Rolls Out Voice Control for Codex Coding Agent on Desktop

September 17, 2026 at 16:13 EDT

Foundation Models
AI Agents
Software Dev & Coding

OpenAI has integrated GPT-Live-1, its full-duplex voice model, directly into Codex on the desktop, enabling developers to control the AI coding agent using spoken commands. The voice layer, powered by GPT-Live-1 and released to the ChatGPT desktop app on July 23, 2026, allows developers to speak tasks rather than type them, creating a hands-free interface for repository management while they focus on higher-level planning.

AI Development Tools

Voice-Controlled Coding Agents Go Mainstream

OpenAI integrates full-duplex voice into Codex, letting developers speak commands to coordinate multiple background agents without typing.

2M+

Codex Weekly Active Users

Weekly Developers (GPT Platform)

10K+

NVIDIA Employees (Codex Access)

$0.05/min

GPT-Live-1 Voice Pricing

Full-Duplex Voice Capability

Traditional Voice

Listen-Then-Respond

GPT-Live-1

Simultaneous I/O

Key advantage: Full-duplex eliminates latency and lost context that plague traditional voice agents, enabling natural interruptions and real-time task coordination.

Platform rollout: ChatGPT Voice integrated into Codex on desktop (July 23, 2026), with support for Mac, Windows, iOS (remote access), and upcoming Android support. Available to Plus, Pro, Business, Edu, and Enterprise users.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.