Google launches Gemini 3.8 Live with real-time video avatars for customer service

Google unveiled Gemini 3.8 Live with Live Avatar on Tuesday, adding near‑real‑time video avatars to its newest live‑dialogue models and positioning them as the next generation of voice‑first customer‑service and sales agents for enterprise clients. The feature, announced alongside Gemini 3.8 Live and Live Extended Thinking, lets developers embed a visual persona that can listen, see and speak, turning a simple voice interaction into a multimodal conversation on a computer or phone screen.

The rollout builds on Google’s push to make its Gemini AI platform more production‑ready for businesses that need reliable, scalable agents. Last month the company highlighted Gemini 3.8 Live as its most advanced voice model, promising “reliable, production‑ready voice agents” for developers and enterprises. By layering a visual avatar on top of that foundation, Google hopes to close the gap between human‑like interaction and the practical demands of tasks such as troubleshooting, product recommendations, or appointment scheduling, where facial cues and lip‑sync can make the exchange feel more natural.

In the demo videos, Google shows both photorealistic and cartoon‑style avatars that match lip movements to speech, display natural facial expressions and manage “fluid turn‑taking” during dialogue. The avatars are powered by Gemini’s reasoning engine, allowing them to trigger background tool calls, fetch data, and keep the conversation flowing without pausing for computation. Enterprise customers can choose from a library of preset avatars or upload high‑quality reference images to create custom personas that reflect brand identity or a specific character. Google says each avatar is watermarked with its SynthID technology and subject to “strict safeguards” designed to protect identity and keep AI‑generated content transparent. Access is limited to companies on an enterprise allowlist, but the company hints that the avatars could soon appear on consumer‑facing interfaces as part of broader customer‑service deployments.

Analysts expect the feature to roll out gradually as more businesses join Google’s Gemini Enterprise program. If widely adopted, the avatars could reshape how companies handle support calls, reducing reliance on human agents while raising questions about user comfort and the quality of AI‑generated interactions. Google’s emphasis on safeguards and branding tools may ease some concerns, but early reactions suggest that users will judge the technology on how well it avoids the “creepy” feel that has plagued earlier AI avatar experiments. The next few months will reveal whether the visual layer delivers the promised boost in engagement or becomes another niche tool for a limited set of corporate users.