OpenAI Rolls Out GPT-Live-1.1, Adding New ChatGPT Voices for Six Languages

OpenAI started rolling out an updated voice model for ChatGPT, called GPT-Live-1.1, on October 15. The release adds two new voices each for French, German, Hindi, Japanese, Korean and Mexican Spanish, along with a new English voice named Bubbie, for a total of 13 new voices.

GPT-Live-1.1 succeeds GPT-Live, the full-duplex voice system OpenAI introduced for ChatGPT in July. Unlike the older, turn-based voice mode, GPT-Live lets the assistant listen and respond at close to the same time, so it can acknowledge what a user is saying with short verbal cues instead of waiting for them to finish speaking.

Alongside the main model, OpenAI also released a smaller variant called GPT-Live-1.1 Quip, which is the default for users on ChatGPT's free tier. The update is also said to cut down on the spoken filler phrases, such as announcing that it is "checking," that the assistant previously used while retrieving information from the web.

The rollout is gradual, so not every account will see the new voices immediately, and availability can vary by region. Users who do have access can change their selected voice from the Voice section of ChatGPT's settings.

OpenAI has not given a timeline for bringing the updated voice model to its API for developers, saying only that access is coming. Several reports also noted early glitches with voice selection and language recognition on Android and iOS following the rollout, which OpenAI has not commented on publicly.

The update extends OpenAI's push to make ChatGPT's spoken interactions feel more natural and more accessible outside English-speaking markets, as it competes with Google's Gemini and other assistants that have also been expanding multilingual voice features this year.