What's this about?
OpenAI added voice to the ChatGPT desktop app on macOS and Windows. You can now control your computer, kick off coding tasks in Codex, and direct several agents running in ChatGPT Work — entirely by talking, like a manager dictating to a team of assistants who fan out and do the work.
What exactly is "full-duplex" voice?
Old voice assistants are walkie-talkies: you speak, stop, then they respond. GPT-Live (the voice model behind this, launched July 8) is a phone call: it listens *while* it talks, so you can interrupt, overlap, and keep the conversation going while it works. When a question needs heavy thinking, it quietly hands the reasoning to GPT-5.5 in the background and keeps chatting without freezing.
What's the community saying?
8.9k likes on the announcement; rolling out to paid tiers globally
- The framing people repeat: this turns "using a computer" into "supervising several workers by voice" — the multi-agent part matters more than the voice part.
- Accessibility angle: hands-free control of real desktop work is a genuine leap for users who can't type comfortably.
- Skeptics recall every "voice control of the desktop" demo since 2010 — the question is whether error rates make it faster than typing for real tasks.