Google announced Live Avatar for Gemini 3.8 Live on September 24, giving its conversational AI an animated face that can respond as you talk. Google Cloud says it’s generally available through Gemini Enterprise. Companies need to be on an approved list to use custom avatars.
Google says the system can listen and process visual input at the same time, speak 97 languages, and keep a conversation going while it uses tools in the background. Those are Google’s descriptions of the product. They don’t tell us how often a person will need to repeat a question or fix a misunderstanding.
For this newsroom, the interesting possibility is letting someone ask questions about a story. If a term doesn’t make sense, a reader could ask for an explanation instead of leaving the page to search for it. Or they could ask which source supports something in the article. That’s an idea to explore, not a feature we’ve added here.
I’d want the answer to stay tied to what I’ve published. If someone asked what I think about a subject I haven’t covered, it should say it doesn’t have that answer. An avatar that invents my opinion would be a problem, even if the conversation felt natural.
People should also know they’re talking to AI. Google says the generated audio and video carry SynthID watermarks. Those aren’t a substitute for telling the person on the other end what they’re using.
The background tools are worth watching too. Answering a question about a subscription is different from changing it. If an avatar can take that next step, the person should see what it’s about to do and confirm it.