Announced September 24, 2026, Live Avatar pairs near-real-time video generation with speech so Gemini 3.8 Live can listen, see, and answer as a lip-synced visual persona — starting in Gemini Enterprise.
2026-09-24 · by Moe Ameen
Google announced Gemini 3.8 Live with Live Avatar on September 24, 2026 — a visual upgrade to its real-time voice model that pairs near-real-time video generation with speech. Where Gemini 3.8 Live (shipped September 15) was a voice that could listen, see, and talk, Live Avatar gives that voice a face: a dynamic on-screen persona that responds with precise lip-syncing, natural facial expressions, and fluid conversational turn-taking. It processes visual and audio input at the same time and answers with expressive audio and video together, in near real time.
Two capabilities carry the experience. The model supports asynchronous tool calling, so it can kick off a lookup or an API call in the background and keep talking while the data comes back — no dead air mid-conversation. And it works across 97 languages, adapting the avatar's lip-sync as it switches. Every audio and video stream carries an imperceptible SynthID watermark so the output stays detectable as AI-generated.
On the avatars themselves, Google is offering a library of preset faces to all enterprise users, while creating a custom avatar from a reference image requires enterprise-allowlisted access. Availability is the headline caveat for creators: Live Avatar launched inside Gemini Enterprise, and Google did not announce a consumer rollout or a price. It's aimed at business use cases — support agents, sales assistants, training — where a talking face on a live conversation is the product itself. That also defines its limit: it's a conversational interface, not a content-production tool. It holds a real-time exchange, but it exports no captioned short, carousel, blog, or scheduled post.
There are two ways to act on this. Topically, it's a live AI story, so drop "Google gave Gemini 3.8 Live a talking face" into [Kompozy](/)'s Quick Ingest and fan one source into a Text Post, an X thread, a brand-exact LinkedIn [Carousel](/glossary/output-buckets), a captioned explainer video, and a blog in your own voice — then schedule the set across the eight social platforms plus blog and email. A single news beat becomes a same-day cross-platform package.
The durable point is about access and output. Live Avatar puts an AI face behind an enterprise paywall and an allowlist, aimed at live support and sales conversations — a face that talks back in the moment and exports nothing you can post. Most creators and small brands can't get it, and even if they could, a real-time chat isn't a content calendar. Kompozy is the opposite build, and it's available today: its AI Influencer persona pool gives you a recurring, face-locked presenter that reads your script to camera. Instead of one ephemeral conversation you get produced, publishable output — [Persona Shorts](/glossary/persona-shorts) and HeyGen avatar video, brand-exact carousels via [HyperFrames](/glossary/hyperframes), photo posts, quote graphics, blogs, and newsletters, all held to one [Persona Brief](/glossary/persona-brief) and fanned everywhere by [Autopilot](/glossary/autopilot). Google built a face for a conversation; Kompozy builds a face for your whole content operation.
It is a visual layer on Google's Gemini 3.8 Live voice model, announced September 24, 2026, that pairs near-real-time video generation with speech. The result is an on-screen persona that listens, sees, and answers with precise lip-syncing, natural expressions, and fluid turn-taking, across 97 languages, with SynthID watermarking on the audio and video.
Not yet. Live Avatar launched inside Gemini Enterprise, and Google did not announce a consumer rollout or a price. All enterprise users get a library of preset avatar faces; creating a custom avatar from a reference image requires enterprise-allowlisted access.
No. Live Avatar is a real-time conversational interface — it holds a spoken, on-screen exchange but exports no captioned short, carousel, blog, newsletter, or scheduled post. To turn a script into a recurring, face-locked persona video and publish it across platforms, you use a content engine like Kompozy.
Google says Live Avatar supports 97 languages and adapts the avatar's lip-sync as it switches between them during a conversation.