Google makes Gemini 3.8 Live Avatar generally available for enterprise

Google Cloud announced on 24 September 2026 that Gemini 3.8 Live with Live Avatar is generally available for enterprise production use. According to the announcement, the Live Avatar feature "can generate video avatars with synchronized lip-syncing," and the custom avatar feature "is available via allowlist only."

Alongside the avatar layer, Google says Gemini 3.8 Live "understands and speaks 97 languages, with automatic language detection," uses native speech-to-speech for interruption recovery, and "executes tools and API calls in the background while continuing the conversation, so the model can acknowledge requests and keep chatting while tasks finish in the background." Live visual understanding can process live camera feeds and screen shares alongside audio at the same time.

Availability covers US and EU endpoints, with provisioned throughput, enterprise compliance and data governance controls, through Gemini Enterprise and the API. Google states that all generated audio and video streams carry imperceptible SynthID watermarks. Gemini 3.8 Live Extended Thinking remains in private preview.

Why it matters for digital humans: This is a hyperscaler shipping the full stack — speech, reasoning, tool calling and a talking face — as one GA product with enterprise contracts attached. For specialist avatar vendors, the differentiator can no longer be "we can put a face on your agent." The allowlist gate on custom avatars is the detail to watch: without your own likeness, the avatar is a stock presenter, which is where most vendors add value. The 97-language figure and the latency behaviour are Google-reported and unaudited; anyone evaluating this against an incumbent should measure time-to-first-frame and interruption handling themselves.