Google adds Live Avatars to Gemini 3.8 Live for enterprise agents

According to Engadget, Google launched Gemini 3.8 Live and Live Extended Thinking last week, billing them as its most advanced live dialogue models and as building blocks for reliable, production-ready voice agents. The new avatar feature adds video to that setup. TechRadar reported the avatars arrive as part of the latest Gemini 3.8 Live model. The two reports differ slightly on how the feature is packaged, but both describe avatars that combine speech, visual presence and enterprise-focused deployment.

Google says the avatars can produce precise lip-syncing, natural expressions and fluid turn-taking, allowing them to look, listen and speak. Engadget quoted Google describing an experience that "listens, sees and speaks with a dynamic visual persona." Because the avatars are backed by Gemini's reasoning, they can trigger tool calls and fetch data in the background while continuing a conversation.

Businesses will be able to choose from a library of preset avatars or create custom ones from high-quality reference images. Engadget reported that customization is meant to preserve "reference likeness, brand styling or character identity." TechRadar reported that businesses could also put a relevant logo on an avatar's uniform to fit company branding. The avatars are only available through enterprise allowlisting, according to Engadget, and are designed for Google's Gemini Enterprise customers. TechRadar reported that the feature is aimed at enterprise users but is more about showing what is possible for customers' end users; possible uses include customer service, receptionists and training.

Safety and transparency are part of the rollout. Google said videos of its avatars include SynthID watermarking to verify that they are AI-generated. "This imperceptible watermark is woven directly into the audio and video output, helping to ensure AI-generated content remains detectable to help minimise misinformation and misattribution," the company wrote, according to TechRadar. Engadget reported that Google described the avatars as built with "strict safeguards designed to respect identity and keep AI-generated content transparent."

In a separate post, Google also said Gemini 3.8 Live can support 97 languages and automatic language switching to make multilingual conversations more seamless, TechRadar reported. The harder-core 3.8 Live Extended Thinking model outperformed GPT-Live-1 Astra and Grok Voice Think Fast 2.0 across multiple benchmarks while costing less to run per hour of input audio, according to the same report.

TechRadar called the avatars "creepy" and said they are meant to make support bots feel more human. Engadget wrote that "like it or not" such avatars may handle customer service or sales support, and noted people's distaste for AI slop. The feature's first practical appearance could be on a computer or phone screen in a customer service or sales role, Engadget reported, though it is currently limited to enterprise customers.