Google has introduced Gemini 3.8 Live with Live Avatar, adding real-time visual presence to its conversational AI for enterprise use. The feature follows last week’s launch of Gemini 3.8 Live and combines its native live dialogue capabilities with low-latency streaming video.
Gemini 3.8 Live with Live Avatar
Live Avatar pairs near-real-time video generation with speech to create a dynamic visual persona that can listen, see, and speak during conversations. It supports precise lip-syncing, natural expressions, and fluid turn-taking.
Multimodal conversations
Live Avatar processes visual and audio inputs simultaneously for enterprise agents. It combines these inputs with the live dialogue model to support conversations involving speech, visual information, and facial expressions.
The feature can support use cases such as customer service and interactive walkthroughs, where an enterprise agent communicates with users through a visual avatar.
Asynchronous tool execution
Live Avatar also uses Gemini’s advanced reasoning with asynchronous tool calling. It can trigger tool calls and fetch data in the background while continuing an active conversation. This allows the agent to handle complex tasks without interrupting the dialogue while it waits for information from a tool.
Multilingual speech-to-speech synchronization
Live Avatar supports native multilingual speech-to-speech synchronization across 97 languages. It dynamically adapts lip-sync and facial expressions when the conversation transitions between languages. Google says the system maintains video fidelity during these transitions without introducing visual drift.
Preset and custom avatars
Organizations can choose from a library of diverse preset avatars or customize their own Live Avatars. Developers can use a high-quality reference image to generate a fully animated and responsive avatar while preserving:
- Reference likeness
- Brand styling
- Character identity
Custom avatar creation is currently available only through enterprise allowlisting.
Safeguards and SynthID watermarking
Live Avatar includes safeguards designed to respect identity and keep AI-generated content transparent. Google says all output generated by its AI products carries a SynthID watermark.
The imperceptible watermark is embedded directly into the generated audio and video, helping make AI-generated content detectable and helping minimize misinformation and misattribution.
Google also provides a model card covering its approach to safety and responsible deployment.
Availability
Gemini 3.8 Live with Live Avatar is available in Gemini Enterprise starting today. Google has also made API documentation available for developers to get started.