OpenAI launches new GPT-Live-1.1 model

OpenAI announced on Thursday, October 15, the release of GPT-Live-1.1, an updated version of its language model that allows ChatGPT to speak in a vocal mode. This update aims to improve the model's ability to interact with users through voice, offering better pronunciation in multiple languages.

Improved language support and new voices

The update includes significantly better pronunciation in several languages, such as French, German, Hindi, Mexican Spanish, Korean, and Japanese. It adds 13 new voices, including two for each of the six languages, plus an English voice named Bubbie. In French, the new voices are called Aube and Rivage.

Full-duplex interaction and responsiveness

GPT-Live-1.1 functions in full-duplex mode, meaning it can listen and speak simultaneously, like a telephone conversation. This allows users to interrupt the model without waiting for it to finish speaking, improving the flow of interaction.

Fewer interruptions and smoother delegation

OpenAI also promises that ChatGPT will repeat less of its "checking..." message when searching the web. Instead, it will use more varied short phrases to announce its searches. The company claims the model’s delegation process—when it hands complex questions to a more powerful model—has become more natural and seamless.

Different models for different users

Not all users will receive the same version of GPT-Live-1.1. It becomes the default model for Plus subscribers, while free accounts will use GPT-Live-1.1 Quip, a lighter version with limited features. The availability of specific voices may vary by region and account, and some users might need to wait for the update to appear in their settings.

API update coming soon

Developers will have to wait for GPT-Live-1.1 to become available in the API, which allows third-party applications to use the model. While no specific date was given, OpenAI stated the update will arrive "soon."