Skip to content
Cyberttak.
Learning & language· 1 min readBreaking

Google DeepMind added live avatars to Gemini 3.8

The new feature allows enterprise users to interact with AI models through real-time video, audio, and fluid facial expressions.

By CyberttakPublished · 09:00
A professional headset with a dark visor and a bright blue ceramic mug sit on a clean studio tabletop
A professional headset with a dark visor and a bright blue ceramic mug sit on a clean studio tabletopCredit: AI-generated · FLUX.2 klein 9B

Google DeepMind released Gemini 3.8 Live with Live Avatar for its enterprise platform. The tool allows AI models to present themselves with a visual persona that mimics human conversation through speech, lip-syncing, and facial expressions.

The system handles audio and visual data simultaneously to enable natural turn-taking. It supports 97 languages with automatic synchronization of lip movements and expressions during multilingual conversations.

Live Avatar brings these capabilities to enterprise agents.

Live Avatar uses asynchronous tool execution, allowing the system to fetch information or perform tasks in the background without pausing the conversation. The company applies SynthID watermarking to all generated audio and video content to signal that it is AI-produced.

Custom avatars are currently limited to enterprise allowlisting. The technology requires high-quality reference images to maintain brand identity and character likeness during animation.

Source: Google DeepMind Blog