- Google has officially announced “Gemini 3.8 Live with Live Avatar,” which combines “Gemini 3.8 Live” with near-real-time animated avatar video.
- By combining the real-time conversational capabilities of “Gemini 3.8 Live” with low-latency video generation, it is a live conversational AI model that delivers a face-to-face-like conversation experience where an expressive avatar listens to the user, observes their reactions, and responds accordingly.
- Use cases such as corporate customer service and interactive operation guidance are anticipated.
Just after announcing “Gemini 3.8 Live” and “Gemini 3.8 Live Extended Thinking”, AI models specialized for voice conversation, Google officially announced “Gemini 3.8 Live with Live Avatar”—which combines “Gemini 3.8 Live” with near-real-time animated avatar video—on Thursday, September 24, 2026 local time.
“Gemini 3.8 Live with Live Avatar” is a live conversational AI model that delivers a face-to-face-like conversation experience by combining the real-time conversational capabilities of “Gemini 3.8 Live” with low-latency video generation, allowing an expressive avatar to listen to users, observe their reactions, and respond. It supports accurate lip-syncing, natural facial expressions, and smooth conversational exchanges.
Intended use cases include corporate customer service and interactive operation guidance. It has begun rolling out on the enterprise AI agent platform “Gemini Enterprise.”
“Gemini 3.8 Live with Live Avatar” simultaneously processes video and audio inputs, understands what it sees and hears in near-real time, and responds with both voice and expressive video.
Additionally, much like “Gemini 3.8 Live,” it allows conversations to continue while tools are called and data is retrieved in the background, enabling tasks that require multiple steps, such as hotel check-in procedures, to be handled without interrupting the conversation.
“Gemini 3.8 Live with Live Avatar” supports 97 languages. Even if the language switches mid-conversation, it adjusts lip-syncing and facial expressions to match the language without degrading video quality or causing unnatural visual misalignments.
“Gemini 3.8 Live with Live Avatar” allows users to choose from a variety of ready-made avatars, or create original avatars based on high-quality reference images while preserving the original appearance, brand style, and character traits. However, custom avatar creation is currently only available to enterprises on Google’s allowlist.
Furthermore, the audio and video generated by “Gemini 3.8 Live with Live Avatar” are embedded with “SynthID,” a digital watermark imperceptible to human eyes and ears.
Source: Google (https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-live-with-live-avatar/)