# Gemini 3.8 Live Gets a Face: Live Avatar Lands in Gemini Enterprise, Custom Avatars by Allowlist Only

Google added Live Avatar to Gemini 3.8 Live on 24 September: a near real-time talking avatar with lip-sync in 97 languages, available in Gemini Enterprise. Custom avatars from your own reference images need enterprise allowlisting, audio and video carry a SynthID watermark, and Google published no price.

By Neomanex. Updated 2026-10-12.

**Our verdict: no change.** [Gemini 3.8 Live](/models/gemini-3-8-live) stays conditional. Google added Live Avatar, a talking face for its real-time voice model, on 24 September, but only inside Gemini Enterprise, with custom avatars behind an allowlist and no published price. If you build voice agents on the Gemini API, nothing changes for you today.

## What happened

Google's 24 September post introduces Gemini 3.8 Live with Live Avatar, which adds a near real-time visual presence to Gemini's conversational AI. It pairs video generation with Gemini 3.8 Live's speech: the avatar lip-syncs, shows expressions and handles turn-taking.

- **Where:** available in Gemini Enterprise from launch day.
- **Languages:** 97, with lip-sync that follows a switch from one language to another mid-conversation.
- **Custom avatars:** organisations can build an avatar from their own reference images, but Google says "custom avatar creation is currently available only through enterprise allowlisting".
- **Watermarking:** SynthID is embedded in both the audio and the video output.
- **Price:** none in the post.
- **Latency:** Google says "low-latency streaming video" but gives no figure.

## Why it matters

A face on a voice agent is the step from call-centre bot to on-screen assistant: kiosks, onboarding videos, support widgets. Google shipping it with SynthID on both tracks is the right default for a product that generates a talking human likeness.

But the launch shape limits who can use it. It lives in Gemini Enterprise, the branded-avatar feature most companies would actually want is gated, and there is no rate card to plan against. Without a price we cannot compare it with the voice-only stack, so it does not move our assessment.

Our Gemini 3.8 Live entry covers the voice model on the Gemini API. It is conditional because Google bills [Gemini 3.8 Live Extended Thinking](/models/gemini-3-8-live-extended-thinking) at the same rate on its [Gemini API pricing page](https://ai.google.dev/gemini-api/docs/pricing), and [Artificial Analysis](https://artificialanalysis.ai/speech-to-speech) scores that sibling higher on its Speech to Speech Index (82.6 against 76.0). An Enterprise-only avatar layer does not change that trade. We will look again when Live Avatar has a published price outside Gemini Enterprise.

## What changes for you

- **Building voice agents on the Gemini API:** nothing. Our pick for new projects remains Gemini 3.8 Live Extended Thinking; choose 3.8 Live only when turn latency beats reasoning depth.
- **Gemini Enterprise customers:** you can try Live Avatar now. If you want a branded avatar from your own images, ask Google about the custom-avatar allowlist.
- **Everyone else:** watch for a published price and availability outside Gemini Enterprise.

Canonical: https://neomanex.com/news/gemini-3-8-live-avatar-gemini-enterprise