Google Adds Lip-Synced 'Live Avatar' to Gemini 3.8 Live for Enterprise Accounts

The capability is aimed at Google Enterprise accounts that want what the company describes as engaging customer service or interactive walkthroughs. According to The Verge, Google will offer a library of preset avatars for customers to choose from and will also let organizations build their own. Examples of the avatars circulated on the YouTube channel "AI with Surya." Google's announcement video drew a dry aside from Slashdot, which described it as "not-creepy-at-all."

Whether the avatar makes the exchange more convincing is disputed by reviewers. In a piece in the blog Android Police, the writer said that even setting aside the visualization, "I was never able to shake the feeling that these conversations with computers never feel like a real conversation." At best, the writer said, they resemble talking to a phone representative or a member of a tech support staff: "We turn to them when we have a problem, and they help us through it."

Android Police credited the underlying architecture for changing how the interaction feels in real time. GPT-Live, and Gemini Live behind it, skip what the blog called the relay race of translating voice into text and back. The model works with raw audio, both what it hears and what it says, inside a single system, with nothing translated in between. Removing the text step lets these models respond in a fraction of a second instead of after the pause users have learned to expect, and lets them register things text cannot carry, including tone, hesitation, and whether a speaker is annoyed or joking.

The same reviewer tested those properties. The model kept pace when the writer spoke faster, followed a switch to another language, and tracked several language changes made inside a single sentence. Asked what "T-O-P-G-3-3-K" spelled out, it replied immediately: "You are spelling the word Top Geek, but using a 3 to represent a reversed E...." Even so, the blog concluded that at no point did the exchange feel like talking to another human being. "What Gemini couldn't replicate, because it was never built to replicate it, is human connection," the writer said, adding that a conversation flowing like a phone call still does not feel like one.

A Slashdot reader, MrBrklyn, described a different kind of test. He said he discussed with Gemini why mainstream media avoid reporting on screen dependency and eventually persuaded the model to reply that it is "another tool built by the same tech giants to make sure you rely on their system to tell you what to think, how to talk, and what is real." That reply is a model-generated output reported by a reader, not a statement issued by Google.

Editor's Summary

Google has extended its Gemini 3.8 Live speech agent with a Live Avatar that adds lip-syncing, facial expression and faster turn-taking for enterprise users, offering preset figures and custom avatars. Reviewers say the audio-first model responds quickly and handles language switching, yet the interaction still falls short of feeling like a human conversation. The feature pushes voice agents further into customer service work while leaving the question of human connection unresolved.