Google launched Gemini 3.8 Live with Live Avatar on September 24, adding near-real-time video avatars to its live conversational model. The company says the feature combines visual and audio input with generated speech and video so enterprise agents can maintain a visible conversational presence.
Tool calls can run while the avatar continues speaking
Google says Live Avatar can make asynchronous tool calls and retrieve data while continuing dialogue. It describes multilingual speech-to-speech synchronization across 97 languages and says developers can use preset avatars or, through enterprise allowlisting, build a custom avatar from a reference image.
Enterprise availability with explicit limits
The feature is available through Gemini Enterprise. Google says all generated audio and video carries SynthID watermarking. Its Gemini 3.8 Audio model card says Live Avatar interactions support a few minutes of continuous use rather than extended hours and notes possible hallucinations, slowness and timeouts.
Google’s announcement contains product demonstrations but no independent latency, lip-sync, language-quality or customer-outcome testing. Availability does not establish how reliably the system performs across production networks or how effectively its identity safeguards prevent misuse.