ChatGPT’s Voice Just Got a Major Upgrade—and It’s Spooky How Human It Sounds

ChatGPT now feels less like a voice assistant and more like someone you know. In a post on X, the company rolled out a major update to Advanced Voice Mode, and it’s one of the most human-sounding AI voices yet.

It adds realistic cadence, soft pauses, emotional cues, and even manages to sound empathetic or sarcastic when it needs to. If you’ve used it before, you’ll immediately notice the upgrade. Think of it as the difference between a monotone audiobook and a friend telling you a story.

Also read: Jony Ive is building AI gadgets for OpenAI—a wearable, smart speaker, and even a robot

Advanced Voice Mode, which runs on OpenAI’s GPT-4o, is now capable of translating languages live in conversation. You can simply ask it to “start translating,” and it’ll keep going until you say stop, replacing the need for most translation apps.

OpenAI says responses come in as little as 232 milliseconds, with an average of 320 ms. That’s close to how quickly humans reply in real life. Voice Mode also got a few improvements earlier this year, including better handling of interruptions and accents.

The updated Voice feature is rolling out now to paid ChatGPT users. OpenAI admits there are still some quirks, like slight pitch shifts or the occasional weird audio blip.