Google's new 'Live Avatar' capability allows Gemini 3.8 Live to present a visual presence — either a cartoon character or a realistic human simulacrum — that lip-syncs and shows facial expressions during conversations, according to blog posts and reports from The Register and The Verge. The feature is currently limited to Gemini Enterprise customers.
In a blog post, Google DeepMind research scientist Shuo-yiin Chang and Gemini software engineer CJ Zheng said the feature processes visual and audio inputs simultaneously to generate richer dialogue. "Live Avatar brings these capabilities to enterprise agents," they wrote, noting the system can transition between the 97 languages it supports "without degrading video fidelity or introducing visual drift."
Avatars can be chosen from a pre-made library, or custom avatars can be deployed when enterprise allow-listing is enabled. The engineers added that the avatars can make tool calls and fetch data in the background, making it suitable for scenarios like handling hotel guest check-ins.
The launch comes amid continued scrutiny of AI humanisation. Five years ago, DeepMind published a paper on the ethical and social risks of language models, warning that anthropomorphising AI could inflate users' estimates of an agent's competence and lead to undue trust. In August 2025, the parents of Adam Raine sued OpenAI, alleging that ChatGPT's "anthropomorphic mannerisms calibrated to convey human-like empathy" contributed to their son's suicide.
More recently, in December 2025, Google researchers published a paper titled "How Tech Workers Contend with Hazards of Humanlikeness in Generative AI," which found that tech workers in focus groups expressed concerns that humanlikeness fosters a false sense of reliability and trust, obscuring errors. Despite these warnings, Google's enterprise deployment of Live Avatar moves ahead, with Chang and Zheng insisting the feature was built with "strict safeguards designed to respect identity, and keep AI-generated content transparent."