Google has made Live Avatar generally available with Gemini 3.8 Live in Gemini Enterprise, adding synchronized generated video personas to its live conversational-agent stack.
The 24 September release is not the launch of Gemini 3.8 Live itself. The underlying real-time conversational model became generally available on 15 September. Live Avatar extends it with synchronized video alongside speech, allowing an agent to respond through a visual persona rather than stopping at text or voice.
Gemini 3.8 Live can process live audio and visual input and execute tools asynchronously while a conversation continues. Google says Live Avatar can transition across 97 languages while adapting lip-sync and facial expressions. It also says generated Live Avatar audio and video carry SynthID watermarks.
Enterprise customers can use preset avatars, while creating customized avatars from reference material remains subject to enterprise allowlisting.
Google presents customer service, interactive walkthroughs, kiosks and other enterprise interfaces as potential applications. These examples are not evidence of adoption: no material production deployment at scale, usage volume, customer preference or commercial outcome has been independently established.
The change is concrete but bounded: generated visual presence is now a deployable interface option for Gemini Enterprise’s live conversational agents. Whether enterprises or their customers will embrace it remains unproven.