Google Debuts Real-Time AI Persona for Conversations with Gemini Live Avatar

Google's Gemini 3.8 Live update has been released, introducing 'Live Avatar', an animated AI persona that responds in real-time during conversations. This feature is currently only available to Gemini Enterprise customers. The avatar can lip-sync and display facial expressions while communicating, supporting up to 97 languages without compromising video quality.

One notable aspect of Live Avatar is its ability to transition between different languages seamlessly. In a shared video by Google, the model is seen speaking English and Japanese simultaneously, with synchronized mouth animation. Additionally, it can retrieve information on-screen while conversing.

The launch of Live Avatar follows closely after the introduction of Gemini 3.8 Live, which processes visual inputs in near real-time. Organizations will be able to create their own avatars using Google's tools, and a library of preset avatars will also be available for selection. However, users can expect a watermark, known as SynthID, to be embedded in the output alongside safeguards designed to respect identity.