Tools 24 September 2 min read
Google releases Gemini Live Avatar for enterprise voice and video
Gemini Enterprise gains animated conversational characters that can respond with speech and video. Custom characters remain restricted to allowlisted organisations.
Businesses can now put a moving, speaking character in front of a Gemini conversation, rather than offering voice alone. Google names customer service and interactive walkthroughs as uses for Gemini 3.8 Live with Live Avatar.
In the announcement, Shuo-yiin Chang and CJ Zheng describe a system that processes audio and visual inputs together and generates near-real-time spoken and video responses. Google claims synchronised lip movements, expressive faces and fluid exchanges between speakers.
A conversation that continues while tools run
The feature can call tools and fetch information in the background while continuing to talk. Google demonstrates this with a hotel check-in, illustrating how an agent might keep a guest informed while handling a task.
Google also claims the avatar can switch among 97 languages without losing video fidelity or lip synchronisation. Those are company performance claims, not independent test results.
Organisations can choose preset characters. Allowlisted enterprise developers can instead generate an animated avatar from a reference image, preserving its likeness or styling. Google says generated audio and video carry its imperceptible SynthID watermark to help identify them as AI-generated.
Teams building spoken customer-service agents can add a visual character and keep dialogue going during data lookups. The access distinction is practical: preset avatars are offered in Gemini Enterprise, while custom identities require allowlisting.
Sources & publication notes
Published in our 25/09/2026 edition. Source dates are shown above.
The AI Street Journal · Free to read
Make this your morning paper.
Three AI stories, clearly explained and illustrated. Free in your inbox.
Prefer to listen? Choose your podcast preferences →