Google Launches Gemini 3.8 Live with Real-Time Avatars

Google launched Gemini 3.8 Live with Live Avatar for Gemini Enterprise, integrating real-time visual personas into its conversational AI. The system utilizes low-latency streaming video generation and speech to enable virtual avatars that can see, listen, and speak simultaneously with natural facial expressions and precise lip-syncing.

Designed for enterprise applications like interactive walkthroughs and customer service, the tool supports asynchronous execution to fetch data without pausing the conversation. The AI can switch between 97 different languages mid-interaction. Organizations have the option to use pre-made avatars or generate custom personas from reference images to maintain brand identity.

To address security and transparency concerns, the company embedded SynthID watermarks into all AI-generated audio and video outputs.