Google brings real-time visual avatars to Gemini 3.8 Live

Google has added Live Avatar to Gemini 3.8 Live, pairing a near-real-time visual persona with its live dialogue models. The feature lets virtual characters listen, see, and speak during conversations, with support for precise lip-syncing, natural expressions, and fluid turn-taking.

According to Google, Live Avatar is powered by Gemini's reasoning capabilities, enabling the virtual character to trigger tool calls and fetch data in the background while a dialogue remains active. The company demonstrated both realistic and cartoon-style avatars in its announcement, showing how they can use facial expressions alongside speech for multimodal interaction.

Google will offer a library of preset avatars, and enterprises can create custom avatars from high-quality reference images to preserve likeness, brand styling, or character identity. Custom avatar access is limited to an enterprise allowlist. Google also said the avatars include safeguards to respect identity and keep AI-generated content transparent, with SynthID watermarks applied to generated output.

Live Avatar is currently aimed at Gemini Enterprise customers. Google said the avatars are designed for use cases such as sales support and customer service, and could appear in customer-facing roles on screens in the future.