Gemini 3.8 Live Adds a Face to the Conversation

Google has added a visual presence to its live dialogue models. Gemini 3.8 Live with Live Avatar, released on 24 September 2026, pairs streaming video generation with speech to produce a persona that listens, sees and speaks in something close to real time.

The parts that are hard to get right are the ones Google lead with: lip-syncing that tracks the speech, expressions that do not sit dead on the face, and turn-taking that does not talk over you. Live Avatar handles visual and audio input at once, and holds speech-to-speech synchronisation across 97 languages without the video quality dropping or the face drifting out of alignment.

Behind the face, tool calls run asynchronously. The avatar can fetch data in the background while the conversation carries on, which is the difference between a system that pauses awkwardly to look something up and one that keeps talking while it does.

Organisations can pick from a library of preset avatars or generate a custom one from a reference image, though custom creation is currently gated behind enterprise allowlisting. Everything the feature produces, audio and video alike, carries a SynthID watermark so it can be identified as AI-generated later.

It is available now in Gemini Enterprise.