Google has launched a system of video avatars capable of holding conversations in 97 languages, moving their lips in sync with speech, and executing internal tasks while continuing to chat with the user. The new feature, called Gemini 3.8 Live with Live Avatar, is now generally available to Gemini Enterprise customers, as confirmed by Google Cloud on September 25.
The launch comes at a time when several major tech companies are competing to dominate automated customer service through AI-generated video. Meta has made moves in the same direction in recent days, positioning conversational avatars as one of the next commercial battles in the sector.
What this Google avatar can do
According to the official Google Cloud announcement, the tool brings together several capabilities designed to replace or complement human agents in customer service:
- Video lip-syncing: the avatar moves its lips in time with what it says, mimicking a real video call conversation.
- 97 languages, with automatic detection of the language spoken by the person on the other end.
- Background task execution: it can activate tools and connect with other applications while continuing to speak, allowing it to confirm a request and continue the conversation without pauses while the process is completed.
- Interruption tolerance: if the user cuts it off mid-sentence, the avatar does not lose the thread of the conversation.
- Live visual perception: it understands what the camera captures and what is shown on a shared screen, in addition to audio.
A real-world example: buying a car with an avatar
Google cites the case of Cox Automotive, which has developed a vehicle-buying assistant based on this avatar for the Autotrader platform. The tool accompanies the user during the search, model comparison, and financing process, pointing out elements on the screen and activating tools while maintaining the conversation.
The system can be integrated into websites, mobile applications, and interactive kiosks, expanding its potential use to sectors such as car dealerships, banks, or physical retail stores.
Verification, service languages, and watermarks
Companies that want to use the tool can choose from a library of avatars already designed by Google. If they prefer to create their own avatar, they must go through an access list and a specific verification process, intended as a security barrier against identity theft or fraudulent use.
The service operates with technical access points in both the United States and the European Union. Furthermore, all generated audio and video content incorporates SynthID, the invisible watermark that Google uses to identify content created by artificial intelligence.
The extended reasoning feature, known as Extended Thinking, remains in private preview for now and is not yet part of the general rollout.
What this means for the end user
For the moment, this is a technology reserved for companies that are already Gemini Enterprise customers, not a feature that an individual or a freelancer can activate on their own today. Its relevance lies in the trend it sets: video customer service, in each person's language and with an assistant capable of resolving real tasks while conversing, is moving from an experiment to concrete commercial products like the one from Autotrader.
For those looking for a virtual presenter for their own videos, there are already more accessible alternatives on the market, such as the HeyGen or Synthesia platforms, aimed at content creators and small businesses.
Frequently Asked Questions
What is Gemini 3.8 Live with Live Avatar?
It is a Google Cloud feature that allows for the creation of AI video avatars capable of holding spoken conversations in 97 languages, with lip-syncing and the ability to execute tasks while dialoguing with the user.
Who can use this tool?
For now, it is available exclusively to companies that are Gemini Enterprise customers, not to individual users or directly to small businesses without that contract.
In how many languages does the avatar work?
It works in 97 languages and automatically detects which language the user is speaking, without the need for prior configuration.
Where can this avatar be used?
It can be integrated into websites, mobile applications, and interactive kiosks, according to Google Cloud in its announcement.
How is it ensured that it is not used to impersonate identities?
Creating a custom avatar requires going through an access list and a Google verification process. Additionally, all generated audio and video carries the invisible SynthID watermark, which allows for identifying that the content was created by AI.
Are there alternatives for creating AI avatars outside of large companies?
Yes, there are platforms like HeyGen or Synthesia, designed for content creators and smaller businesses that want to generate videos with virtual presenters in Spanish or other languages.
