Gemini gets a live face. A business agent can use tools in the background while talking

Google is adding a generated visual presence to its live voice model Gemini 3.8. Live Avatar combines near-instant video generation with speech, visual input and the ability to use additional software tools. The result is not merely an animated head attached to a chatbot. The agent continuously processes audio and video, generates speech and synchronizes lip movement, facial expressions and other visual behavior with it. Google has offered the technology in Gemini Enterprise since September 24. For businesses, the ability to use tools asynchronously is important. Google demonstrates a situation in which an avatar talks with a hotel guest while performing the steps needed for check-in in the background. The conversation does not have to pause while a database query or another operation finishes. This brings the voice agent closer to real service work. A human receptionist also continues talking while working with a reservation system. Existing bots often feel unnatural precisely because there is a pause after every request. The system supports switching among 97 languages and, according to Google, continuously adapts lip synchronization and facial expression. For global customer service, that is economically interesting: one digital worker can serve many language markets without separate filming or dubbing. Companies can also create custom avatars from a reference image, although this feature currently requires special approval. This is where the more sensitive issue of identity begins. The more convincing a digital character becomes, the more important it is that the customer knows they are not communicating with a human. Google says generated audio and video are marked with SynthID. An invisible digital marker can help determine later that the content was created by AI. For live customer communication, however, a hidden watermark alone should not be sufficient. A transparent disclosure must be understandable to the user directly. Another question is how to measure the benefit. A realistic avatar is visually impressive, but compared with a voice-only agent it consumes additional compute. A company must determine whether the visual component actually improves service success, customer trust or accessibility, for example for users who rely on lip-reading. Identity substitution is another risk. If a convincing digital character can be created from a single image, a company needs a strict approval process governing who may be represented this way and where the avatar may be used. The technology may therefore be useful in hospitality, education, retail or technical support. It will not make economic sense everywhere. In many processes, voice or text will be cheaper and clearer. KEY TERMS Live avatar: An AI-generated audiovisual character responding in near real time. Asynchronous tool use: A situation in which an agent starts an external operation and continues another activity while it runs. SynthID: Google technology for machine-detectable marking of AI-generated content.

Gemini gets a live face. A business agent can use tools in the background while talking

The full article is available to users registered at Hard Skills