Gemini Live’s native audio model helps it speak more like a human

[phillynews215]

TL;DR

• Google previewed Gemini Live model updates at the Google Pixel 10 launch event, and they’re now rolling out widely for iOS and Android users.
• Gemini Live is now likely powered by Gemini 2.5 Flash Live API native audio, making for more “adaptive and expressive” chats.
• Users can personalize Gemini Live’s speech to match their preferences, changing aspects of its delivery like speed, tone, and accents.

Gemini Live is Google’s conversational AI assistant, and the company is rolling out a model update that aims to make it more natural and human-like. Google first previewed the improved experience, known as native audio output, at the Google Pixel 10 launch event. Now, it’s available to all Gemini Live users starting today following a limited rollout in October.

Google explains that Gemini Live’s new model updates “dramatically improve its ability to understand and use key elements of human speech.” This results in “more adaptive and expressive conversations,” according to the blog post. Gemini Live users can take advantage of the assistant’s better understanding of speech and language to tailor their experience. For example, users might ask Gemini Live to speed up, slow down, or adjust its tone and accent.




More...