Google gives Gemini Live a real-time talking face in 97 languages

Google released Gemini 3.8 Live with Live Avatar on 24 September 2026, making it generally available the same day in Gemini Enterprise. The feature adds near real-time generated video to the speech of Google’s live dialogue model, so a conversational agent answers with a visible face that lip-syncs, changes expression and takes turns in conversation.

Live Avatar supports 97 languages and keeps the lip-sync aligned when a speaker switches language mid-conversation. Google pitches it for enterprise uses such as customer service, hotel check-in and guided product walkthroughs. Companies can pick from a library of prebuilt avatars; building a custom avatar is available only through enterprise allowlisting.

Every audio and video stream the system produces carries SynthID, Google’s imperceptible watermark, so that the output can be identified as AI-generated. The blog post is signed by Shuo-yiin Chang and CJ Zheng of the Gemini Audio team, who describe the release as a way to give AI agents a visual presence in real-time conversation.