Gemini 3.8 Live Avatar ships lip-synced video on US and EU endpoints

Google Cloud has made Gemini 3.8 Live with Live Avatar generally available in Gemini Enterprise as of September 24, 2026. First previewed at Google Cloud Next 2026, it is now billed as ready for enterprise production: a talking face for conversational video agents across web, mobile, and interactive kiosks, served from US and EU endpoints.

This is the visual layer on Gemini 3.8 Live’s native speech-to-speech foundation. Google’s research blog says Live Avatar couples live dialogue with low-latency streaming video so agents can listen, see, and speak with a dynamic persona—precise lip-sync, natural expressions, and fluid turn-taking. The Verge reported the lip-synced, expressive face and that access is currently limited to Gemini Enterprise customers.

Teams can start from a library of curated, pre-built avatars. Custom identities, generated from a high-quality reference image while preserving likeness and brand styling, stay behind a strict enterprise allowlist and verification process. Google Cloud is explicit: custom avatars are allowlist-only.

Google says Gemini 3.8 Live understands and speaks 97 languages with automatic detection, and that Live Avatar can switch languages mid-conversation while adapting lip-sync and expression without degrading video fidelity or introducing visual drift. All generated audio and video streams carry imperceptible SynthID watermarks. The Cloud post also lists provisioned throughput, enterprise compliance, and strict data governance on those US and EU endpoints.

Gemini 3.8 Live Extended Thinking remains in private preview. Today’s GA is the production talking-face path, not last week’s Live model launch.

If you already have Gemini Enterprise, try Live Avatar in the multimodal live studio and review the Gemini Live API documentation. For provisioned throughput or custom-avatar allowlisting, Google Cloud says to contact your sales representative.

Sources