ChatGPT's voice gets new voices and better accents

What's new

On October 15, 2026, OpenAI announced GPT-Live-1.1. It is an update to the GPT-Live voice model, which launched in July 2026. The update adds new voices for French, German, Hindi, Mexican Spanish, Korean and Japanese. There is also a new English voice called Bubbie.

According to a guide from The Rundown AI, the voices include Aube and Rivage (French), Sonna and Miko (German), Leher and Neer (Hindi), Himari and Rindō (Japanese), Haetsal and Hanul (Korean) and Tinta and Arrullo (Mexican Spanish). The exact number of new voices is unclear in the sources. Which voices you can choose varies by account and region.

The rollout has begun in Voice in ChatGPT, Codex and Dots, and is reaching ChatGPT users around the world gradually.

What's smart

According to the coverage, the model is better at accents, intonation and vocal expression across several languages. That means the voice should sound more natural when speaking languages other than English.

A small but noticeable detail: when ChatGPT looks something up, it no longer says "checking" over and over. Instead, it uses varied phrases while it searches for information.

The model has what is known as a full duplex design. That means it can listen and talk at the same time, just like in an ordinary conversation where people interrupt each other. This also makes it useful for practicing foreign languages.

Cheaper or better?

Free and Go users get the smaller model, GPT-Live-1.1 Quip, while Plus and Pro users get the full GPT-Live-1.1. Free users can access the voice feature via the voice icon in ChatGPT. No new pricing has been announced for regular users.

For businesses and developers, the picture is less clear. API access to GPT-Live-1.1 is listed as "coming soon," with no date or price. OpenAI's API documentation mentions only gpt-live-1, and the known API price of $0.05 per minute applies to that model. OpenAI has also not published accuracy scores for individual languages, response times or full technical specifications for 1.1. Comparisons with competitors' voice features are not included in the sources, and it is not stated whether anything applies specifically to Denmark or the EU beyond the fact that the rollout is global.

Be aware of the terms: according to OpenAI's terms of service, output from ChatGPT Voice is limited to non-commercial use. It may not be distributed as standalone audio files or used in podcasts without permission.

What it's good for

The update is especially interesting for people who talk to ChatGPT in languages other than English. If you are learning French, German, Japanese, Korean or Spanish, you can practice pronunciation and conversation with a voice that should sound more natural and can listen while it speaks.

It is also good for everyday use when you want answers without having to type, for example when you ask about something on the go and let the voice find the information. Businesses that want to build their own solutions on the model will have to wait for API access to arrive.