ChatGPT Voice Desktop Launch: GPT-Live, Appshots, and the Unified Super App

GPT-Live Technology: Replacing Advanced Voice Mode

GPT-Live is a new family of voice models built on a full-duplex architecture — it can listen and speak simultaneously, rather than alternating in a turn-based "push-to-talk" style . Two versions were released on July 8, 2026:

• GPT-Live-1 — for paid subscribers (Go, Plus, Pro)
• GPT-Live-1 mini — for Free-tier users

This system replaces the prior Advanced Voice Mode entirely . The key architectural changes are:

• The model processes audio input and generates audio output continuously, not through rapid turn-switching
• It makes decisions about whether to answer, pause, or invoke a tool "many times per second"
• It delegates complex reasoning tasks to GPT-5.5 (and later GPT-5.6) in the background
• It supports natural interruptions, overlapping speech, and real-time backchanneling ("uh huh")
• It can display information on the screen while conversing

The prior Advanced Voice Mode (launched in mid-2024) was a turn-based system that alternated between listening and speaking. It was fully replaced by GPT-Live on July 8, 2026 . The old macOS desktop voice feature had been retired even earlier — on January 15, 2026 — as part of a broader plan to streamline and enhance voice capabilities ahead of the GPT-Live rollout .

Hands-Free Coordination Across Chat, Work, and Codex

ChatGPT Voice on desktop lets users control the entire unified app by voice across three modes — Chat, Work, and Codex — all accessible from a single desktop shell .

• Chat — the standard conversational ChatGPT interface
• ChatGPT Work — launched July 9, 2026, a new autonomous agent for multi-step professional tasks (creating sheets, slides, docs, web apps) powered by GPT-5.6 and Codex
• Codex — the dedicated coding agent, now integrated into the same desktop app

Users can speak commands to direct multiple agents running in Work or Codex simultaneously, for example: "Check my calendar, draft an email, and prepare a meeting summary" .

Appshots Feature (macOS Only)

On the Mac desktop app, ChatGPT Voice integrates with Appshots, a feature originally released for Codex in May 2026 that lets the AI capture the frontmost app window — including its screenshot and accessible text — for context .

• How it works in Voice mode: Enable "Screen context" in Settings > Voice, then say "Take a look at this." ChatGPT captures the window's image and text, including content outside the visible area
• How it works in Codex: Press both Command keys to send the frontmost window to Codex
• This gives ChatGPT real-time visual and textual context from design tools, dashboards, documents, and other apps without requiring the user to copy, paste, or describe what is on screen

Integration with Computer Use and Plugins

ChatGPT Voice on desktop leverages Computer Use (OpenAI's capability to control the OS), local files, and plugins to execute spoken requests . The system can:

• Navigate the operating system, check calendars, open applications, and interact with on-screen elements
• Use plugins for extended tool access
• On Windows, it provides the same computer-control capabilities minus the Appshots screen-context feature

Broader Consolidation: Codex + ChatGPT + Work → One GPT-5.6-Powered Desktop App

ChatGPT Voice's desktop launch sits on top of a major product consolidation that happened two weeks earlier, on July 9, 2026:

• What happened: The standalone Codex app was merged into the new ChatGPT desktop app. The combined app contains three modes — Chat, Codex, and Work — running in one shell on macOS and Windows
• ChatGPT Work is billed as a combination of ChatGPT and Codex, giving non-technical users access to Codex's capabilities for document creation, workflow automation, and project management
• OpenAI also shipped a new Chrome extension, revamped in-app browser, and faster Computer Use powered by GPT-5.6