ChatGPT’s voice just got smarter and more human: All the details

ChatGPT

OpenAI is rolling out a major upgrade to its Advanced Voice feature in ChatGPT for paid users, bringing a noticeable leap in how natural and expressive its AI sounds. The update refines speech to be more fluid, emotionally intelligent, and intuitive. Think fewer robotic cadences and more realistic pauses, tone shifts, and even the occasional splash of sarcasm or empathy, said OpenAI in a blog post.

This isn’t just about sounding better, though, says OpenAI. The update also introduces real-time translation that makes cross-language conversations seamless. You can now ask the AI to translate between languages—and it will keep doing so until you tell it otherwise. That means a full conversation at a café in Paris or a business call with a colleague in Seoul can flow without needing to constantly repeat translation requests. It’s conversational interpreting, on-demand.

The upgrade is available now to all paying users on all platforms. Just tap the Voice icon and talk. Earlier this year, OpenAI also made tweaks to reduce interruptions and improve accent handling, adding to the overall polish.

Still, it’s not perfect. OpenAI admits the upgrade may occasionally affect audio quality, with some voices experiencing pitch or tone inconsistencies. There’s also a rare but known glitch where Voice Mode can hallucinate unexpected audio—like gibberish or even phantom background music. The company says it’s actively working to fix those issues.