OpenAI's GPT-Live: Talking to ChatGPT Now Feels Much More Like a Real Conversation
OpenAI is bringing voice chat with ChatGPT closer to a real conversation. Its new voice model, GPT-Live, runs on a "full-duplex" architecture — meaning it can listen and speak at the same time. It can wait without cutting you off, show it's following with cues like "mm-hm," and jump into fast back-and-forth. When a hard question comes up, it hands the task to a stronger model in the background (GPT-5.5 for now) and keeps talking to you while the answer is prepared.
In brief
- What changed: ChatGPT Voice now runs on GPT-Live. The old system worked turn-by-turn (you stop, it speaks); GPT-Live is continuous and has a better sense of timing. Nine voices were reworked.
- Intelligence in the background: when a task needs search, reasoning, or agentic work, GPT-Live hands it to GPT-5.5 and keeps the conversation going while it works. You can pick the reasoning level — Instant, Medium, or High.
- The new experience: it listens better (focuses on your voice over background noise, waits while you think) and can show visual cards for weather, stocks, or sports as you talk. It supports search, memory, images, and file uploads.
- Access: live today worldwide on iOS, Android, and ChatGPT.com. GPT-Live-1 is the default for Go/Plus/Pro; GPT-Live-1 mini is the default for Free. No video or screen-share in voice yet (coming soon); the API is coming "soon."
Our take
The value for you: you can now hold an actual conversation with AI instead of "command, then wait" — and hands-free. Testing an idea, getting a quick answer, or dictating notes while driving, walking, or with your hands full gets far smoother. The most useful part is that it pushes the heavy work to the background: while you keep talking, the model searches and reasons and gets the answer ready. And it's on even in the free tier (as mini), so trying it needs no subscription. In short, voice AI moves from a novelty to something that fits into daily workflow — you can start handling a lot of your work by voice instead of a keyboard.
Kaynak: OpenAI