Google has added Live Avatar to Gemini 3.8 Live, giving enterprise customers an animated face that speaks and responds during a conversation. The feature is available in Gemini Enterprise. Google says the avatar can listen, process visual input, speak, and fetch information through tools while the conversation continues.
The new part is the visual presentation around Gemini Live. Google launched the underlying 3.8 Live model last week. This release adds synchronized facial expressions and lip movement to its spoken responses. Google says the avatar can switch among 97 languages during a conversation. The company shows that behavior in demos, but has not published a measured latency figure or a comparison with a voice-only agent in its announcement.
Access and identity controls
Organizations can choose from preset avatars. Google also says developers can create one from a reference image, although custom avatar creation currently requires enterprise allowlisting. Google says generated audio and video carry its SynthID watermark and that it has safeguards intended to protect identity. The announcement does not describe a public test of how reliably those safeguards stop impersonation.
The feature could make an agent easier to follow in customer service or guided tasks, especially when it displays information while speaking. A face also changes what users may infer about the person or organization behind a response. That makes identity and disclosure choices part of deployment, alongside model quality.
