Google adds animated faces to Gemini 3.8 Live

The feature, currently restricted to Gemini Enterprise customers, integrates facial expressions and on-screen information displays directly into the conversational interface. Google demonstrated the avatar’s capabilities by switching between English and Japanese, showing precise mouth movements that align with the specific phonetics of each language. This follows the recent debut of the Gemini 3.8 Live model, which emphasizes near real-time processing of visual inputs.

Organizations gain the ability to deploy custom avatars from a preset library or build their own proprietary versions. To address concerns regarding authenticity and security, Google confirmed that all avatar output is embedded with its invisible SynthID watermark. These digital personas also include built-in safeguards designed to respect identity, ensuring the visual output remains tethered to organizational standards while the model handles complex, multilingual interactions.