Google has launched Live Avatar, a new feature for its Gemini 3.8 Live model. It puts an animated AI persona on screen that responds to you in real time. According to The Verge AI, the avatar lip-syncs, changes facial expressions as the conversation moves along and can bring up information on screen while it talks. For now, only Gemini Enterprise customers can use it.
The launch comes about a week after Google revealed Gemini 3.8 Live itself. Google says that model “processes visual inputs in near real-time.” Put the two together and Gemini can now see, listen, talk and show a face, all in one live conversation.
What Live Avatar Brings to the Table
- Real-time lip-sync and facial expressions – The avatar doesn’t just play audio over a still image. Its mouth moves in time with what it’s saying, and its expressions change during the conversation. That’s the difference between a talking picture and something that feels like a real exchange.
- Support for 97 languages without visual glitches – This is the most impressive technical claim. Google says Live Avatar can switch between its 97 supported languages “without degrading video fidelity or introducing visual drift.” A demo video shows the avatar speaking English and Japanese, and its mouth movements match the words in both. Anyone who has watched badly dubbed video knows how hard this is to get right.
- On-screen information while it speaks – The avatar can show supporting information on screen during the conversation. For a support agent or product guide, that matters. You get a face explaining something and the visual details to back it up.
- Preset avatars or your own custom persona – Google will offer a library of ready-made avatars for customers to choose from. Organizations can also build their own. That means a bank, a retailer or a training company could give its AI assistant a branded face instead of a generic one.
- SynthID watermarking and identity safeguards – Everything Live Avatar produces carries Google’s invisible SynthID watermark, which marks the content as AI-generated. Google also says it has safeguards to “respect identity.” That’s a clear nod to deepfake worries, especially since organizations can now create custom faces.
Who Can Use It Right Now
Only Gemini Enterprise customers have access. Google hasn’t said anything about a consumer rollout, and the source doesn’t include pricing details. That tells you a lot about the strategy. Google is aiming Live Avatar at businesses that want customer-facing AI with more personality than a chat window.
Why This Matters
What stands out here is the target. AI avatars aren’t new. Companies like Synthesia and HeyGen have built whole businesses around AI video presenters. But most of those tools create pre-recorded videos. Live Avatar runs in real time and is built directly into a frontier model’s conversation loop.
That opens up some obvious use cases:
- Customer support: A face-to-face help agent that speaks the customer’s language
- Multilingual service: A single avatar that switches languages mid-conversation for global audiences
- Training and onboarding: Interactive guides that explain things and show visuals while they talk
- Branded assistants: Custom personas that match a company’s identity
The 97-language support could be the feature that sells it. For global companies, one avatar that works across markets is much simpler than localizing video content region by region.
The Caveats
There are a few things to keep in mind. Access is enterprise-only for now, so most people won’t try it anytime soon. The demos come from Google, and real-world performance on messy customer conversations still has to prove itself. Custom avatars also raise tough questions. SynthID watermarks help, but they’re invisible, so the person on the other end may not know for sure they’re talking to AI unless the company says so.
What Comes Next
Google is betting that people will engage more with AI that has a face. If enterprise customers go for it, a consumer version inside the Gemini app seems like a logical next step. Expect OpenAI, Microsoft and others to answer with their own real-time avatar features, too. The Verge AI’s original report has the demo footage and more details.