Google has introduced the Live Avatar feature for its Gemini 3.8 Live model, designed to provide real-time visual interaction for enterprise applications. The system combines video generation with voice processing to enable dynamic communication through precise lip-syncing and natural facial expressions.
The technology leverages Gemini's reasoning capabilities to perform asynchronous tool calls. This allows the virtual agent to retrieve data and process complex tasks in the background without interrupting the ongoing conversation flow.
Live Avatar supports native synchronization across 97 languages, maintaining visual fidelity during seamless language switching. Enterprises can utilize a preset library of avatars or create custom versions based on high-quality reference images to match specific brand styles.
To ensure content authenticity, Google integrates SynthID invisible watermarking into both audio and video outputs. This security measure is designed to identify AI-generated content and mitigate risks associated with misinformation or incorrect attribution.
The feature is currently available through Gemini Enterprise. Developers can access the functionality and begin integration by utilizing the official documentation and API provided by the company.
