Google has introduced a new way to talk to AI. The feature is called Gemini 3.8 Live with Live Avatar. It adds a real-time animated face to voice conversations. Google made the announcement on September 24, 2026, according to the company’s official blog.
The launch builds on last week’s Gemini 3.8 Live release. That earlier model focused purely on voice dialogue. Live Avatar now pairs that voice engine with generated video. The result is a persona that listens, sees, and speaks at once.
How Gemini 3.8 Live Avatar Works Across 97 Languages
The system processes visual and audio inputs simultaneously, according to Google. It can view a live camera feed. It can also see a user’s screen share. Lip-syncing and facial expressions stay accurate throughout. The avatar switches languages mid-conversation without losing sync.
Google says the feature supports 97 languages in total. It automatically detects speech changes and adjusts video alignment. There is no visible delay or drift when languages change. This matters for global support teams and multinational customer service.
Asynchronous Tool Calls Keep Conversations Flowing
Live Avatar can also run background tasks while talking. Google calls this feature asynchronous tool execution. The avatar can query backend systems without pausing speech. One example Google gave involves checking a hotel guest in.
In an insurance demo, a user shows damage on camera. The AI checks policy details while still talking. It verifies rules and builds an adjuster packet. The conversation never stops during these background steps.
Custom Avatars and SynthID Watermarking
Businesses can pick from a library of preset avatars. They can also build a custom one instead. Developers upload a single reference portrait through the Agent Development Kit.
Google then generates a fully animated, responsive persona from that image.
Custom avatar creation is currently limited to enterprise allowlisting, Google confirmed. Every output carries an invisible SynthID watermark.
This applies to both the audio and video streams. The mark helps prevent misinformation, per Google’s developer documentation.
Why Gemini 3.8 Live Avatar Matters for Enterprise AI
Gemini 3.8 Live with Live Avatar is now generally available. It runs inside Gemini Enterprise, Google’s business AI platform. Access currently spans US and EU endpoints. Enterprise compliance and data governance controls are included.
The launch follows Google’s Gemini 3.8 Flash release earlier this month. It also arrives as rivals push agentic tools of their own. One example is Meta’s Muse agent, launched recently.
Android Headlines reports Google’s stock fell roughly 20% recently. That drop followed reported delays with Gemini 3.5 Pro. DeepMind’s Koray Kavukcuoglu has also teased an early Gemini 4 launch.
For now, Live Avatar signals where enterprise AI is heading. Voice alone may no longer be enough. A visible, responsive face could become standard for AI agents. Businesses adopting the feature gain a more human-like interface.
