Google gives Gemini 3.8 Live a face and expressions

Google is adding Live Avatar, an animated on-screen persona, to Gemini 3.8 Live for Gemini Enterprise customers. The feature lets users speak with the AI while an avatar lip-syncs its answers and changes facial expressions in real time.

Emma Roth reports for The Verge that the feature is currently limited to Google’s enterprise offering. Organizations can select from a library of preset avatars and will also be able to create their own versions.

Live Avatar builds on Gemini 3.8 Live, which Google introduced recently as a model designed to process visual inputs in near real time. The new interface adds a visible speaker to those conversations. It can also display information on screen while responding, which could make it useful for guided explanations, presentations, and customer-facing interactions.

Google says the avatar supports transitions among 97 languages without reducing video quality or causing visible inconsistencies in the character’s appearance. In a Google demonstration, the same avatar speaks English and Japanese, with mouth movements synchronized to both languages.

Identity safeguards and watermarking

The company says output produced through Live Avatar includes an invisible SynthID watermark. SynthID is Google’s system for identifying AI-generated content. Google also says the feature includes safeguards intended to respect identity, although it has not detailed how those protections will work for customer-created avatars.

The enterprise-only rollout places the technology first with organizations that may use it in internal communications, support tools, training, or branded conversational experiences. Google has not announced a wider consumer release.