Google has introduced Gemini 3.8 Live with Live Avatar, a new feature that adds near real-time video to its conversational AI. The company announced the launch on September 24, 2026.
The release comes one week after Google launched Gemini 3.8 Live. The new avatar feature is available through Gemini Enterprise, the company's AI platform for businesses.
Google is owned by Alphabet, its parent company. The announcement was written by Google research scientist Shuo-yiin Chang and software engineer CJ Zheng on behalf of the Gemini Audio Team.
Live Avatar pairs speech with near real-time video generation. The result is an AI agent that can listen, see, and speak through an animated on-screen character.
How Live Avatar Works
Google said the feature offers precise lip-syncing, natural facial expressions, and smooth turn-taking. These are meant to make conversations with AI feel more natural and easier to follow.
The system processes visual and audio inputs at the same time. It then responds with both expressive audio and video.
Google said businesses could use the tool for customer service or interactive walkthroughs. In one demo, an avatar checked in a guest at a hotel.
The feature also supports asynchronous tool calling. This means the avatar can fetch data or run tasks in the background while it keeps talking with the user.
Live Avatar supports speech in 97 languages. Google said the avatar can switch languages in the middle of a conversation.
When it switches, the lip-sync and expressions adjust to match the new language. Google said this happens without lowering video quality or causing visual drift.
Custom Avatars and Safety Measures
Companies can pick from a library of preset avatars. Each one has its own look, voice, and expressive style.
Organizations can also create custom avatars to match their brand. Developers can use a high-quality reference image to generate a fully animated, responsive avatar.
Google said the custom avatar keeps the likeness, styling, or character identity of the original image. For now, custom avatar creation is only available through enterprise allowlisting.
On safety, Google said it built Live Avatar with strict safeguards designed to respect identity. All audio and video output from the feature is watermarked with SynthID.
SynthID is an invisible watermark woven directly into the content. Google said it helps keep AI-generated material detectable and aims to reduce misinformation and misattribution.
The company pointed readers to its model card for more details on its approach to safety and responsible deployment.
Gemini 3.8 Live with Live Avatar is now available in Gemini Enterprise. Google said developers can explore its API documentation to get started.
