Google’s Gemini 3.8 Live adds a lip‑synced Live Avatar for enterprise agents

Google updated Gemini 3.8 Live to include Live Avatar, an animated, lip‑synced visual persona that responds in near real time during conversations. The avatar moves its mouth to match spoken output and shows different facial expressions while the model speaks. Google demonstrates the avatar switching languages mid‑conversation without visual drift.

Live Avatar pairs Gemini 3.8 Live's low‑latency conversational model with real‑time video rendering. Key capabilities shown and described in coverage and Google posts:

• Synchronized lip movements that match the generated audio.
• Facial expressions to convey basic emotional or conversational cues.
• Seamless transitions across 97 supported languages with preserved video fidelity.
• On‑screen display of contextual data, diagrams, or reference material while the avatar speaks.

Live Avatar is being released to Gemini Enterprise customers. Google's Cloud blog and product announcements mark the feature as generally available for enterprise production after previewing it at Google Cloud Next 2026.

Google frames the Live Avatar for agent‑style workflows such as customer support and other enterprise conversations. The avatar can act as a visual presence for conversational agents and surface relevant information during exchanges, which targets scenarios where a visual cue and contextual displays could speed understanding.

Tech outlets reported the launch with attention to the technical parts (lip‑sync, multilingual sync across 97 languages, on‑screen data). Some coverage highlights potential benefits for making automated agents feel more human, while other outlets described the avatars as "creepy," reflecting mixed user sentiment about realistic animated agents.

Practical considerations for enterprises

• Deployment: Live Avatar is available through Gemini Enterprise, so organizations already on that tier can evaluate it as part of agent workflows.
• Languages: Support for 97 languages means global teams can use a single avatar framework across many regions without separate rendering pipelines.

• Perception risks: Expect testing for user acceptance—some audiences may find realistic animated agents off‑putting rather than reassuring.

Enterprises evaluating Live Avatar should test it in real customer scenarios, measure comprehension and satisfaction, and monitor any negative reactions tied to avatar realism. Watch for announcements about integrations with connected apps and agent toolchains that bring contextual data into the avatar display.