An enterprise avatar for a live conversation
Google made Gemini 3.8 Live with Live Avatar available in Gemini Enterprise on September 24, following the September 15 release of its Gemini 3.8 Live speech models. The new feature pairs live speech with generated streaming video, letting an agent listen and view input while returning synchronized audio and an animated face. TechRadar reported the enterprise availability the following day. [1][2]
Google positions the feature for customer-facing work such as service conversations and interactive walkthroughs. The avatar can run tool calls in the background while dialogue continues. Google says its speech-to-speech synchronization can switch among 97 languages during a conversation. Those behaviors are described and demonstrated by Google; the coverage does not independently measure latency, language quality or task success. [1][2]
Preset avatars now, custom creation by allowlist
Enterprise customers can select from preset avatars. Google says creating an avatar from a reference image is available only through enterprise allowlisting. Google also says it embeds SynthID watermarks in generated audio and video. The announcement does not establish general consumer availability. [1]
Atlas interpretation: This release extends an already launched live dialogue model with a generated visual presence for enterprise agents. Its audiovisual variant arrived separately from the September 15 Gemini 3.8 Live models, which remain their own Atlas event. [1]
Sources
- Introducing Gemini 3.8 Live with Live Avatar
Google · Sep 24, 2026
- Google's creepy new Gemini Live Avatars want to try and make online support bots feel more human
TechRadar · Sep 25, 2026