Bringing a Real-Time Visual Presence to Conversational AI

Google DeepMind has introduced Gemini 3.8 Live with Live Avatar, adding a near real-time visual presence to its conversational AI models. Building on the recent rollout of Gemini 3.8 Live, the new feature natively couples live dialogue capabilities with low-latency streaming video. According to the official announcement published by Google DeepMind, this pairing creates an interactive persona that listens, sees, and speaks dynamically.

As noted in reporting from The Verge, the update gives Google's AI a visible face that can lip-sync and display various facial expressions during conversations. The feature is currently available exclusively to enterprise users through Gemini Enterprise, transforming standard digital exchanges into more interactive and accessible customer service experiences or virtual walkthroughs.

Multilingual Capabilities and Seamless Code-Switching

Conversation relies on multiple modalities, including audio, sight, and facial expressions. The Live Avatar processes visual and audio inputs simultaneously to foster comprehensive interactions. Furthermore, the system incorporates native multilingual speech-to-speech synchronization, enabling it to dynamically adapt its lip-syncing and expressions across 97 different languages.

Google highlights that the model can seamlessly transition across these languages mid-conversation without degrading video fidelity or introducing visual drift. Demonstrations show the avatar shifting smoothly between languages like English and Japanese while maintaining accurate mouth animations.

Advanced Reasoning and Asynchronous Tool Execution

Beyond visual animation, Gemini 3.8 Live with Live Avatar is supported by advanced reasoning capabilities. The platform features asynchronous tool calling, allowing the AI to trigger tool calls and fetch data in the background while active dialogue continues uninterrupted.

This functionality equips enterprise agents to manage complex multi-step tasks, such as checking in a guest at a hotel or pulling up information on-screen, without breaking the natural flow of the conversation.

Customizable Brand Identities and Enterprise Availability

Organizations frequently require distinct visual identities to align with their brand standards. While Google offers a diverse preset library of characters, developers can also create custom Live Avatars. By starting from a high-quality reference image, developers can generate a fully animated, responsive avatar while preserving the original reference likeness, brand styling, or character identity.

At launch, custom avatar creation is restricted and available exclusively through enterprise allowlisting. Interested developers and enterprise customers can access the feature via Gemini Enterprise and review the developer documentation to get started.

Safety, Watermarking, and Responsible Deployment

To address concerns regarding identity preservation and transparency, Google designed Live Avatar with strict safeguards. All AI-generated audio and video outputs are embedded with an invisible SynthID watermark woven directly into the media streams.

This watermark ensures that AI-generated content remains detectable, helping to minimize misinformation and misattribution. Detailed safety practices and deployment strategies are further outlined in the official model card.

Sources

• Google DeepMindIntroducing Gemini 3.8 Live with Live Avatar
• The VergeGemini 3.8 Live with Live Avatar gives Google’s AI a face