Google Makes Gemini 3.8 Live Avatar Available in Gemini Enterprise
The September 24 release adds generated video personas to Google's live dialogue model, but custom avatars remain allowlisted and performance claims are still first-party.
Edited by Tyronne Panaino
Google made Gemini 3.8 Live with Live Avatar available in Gemini Enterprise on September 24, 2026. The feature couples Google's live dialogue model with generated video so an enterprise agent can listen, see, speak and appear as an animated persona during the same continuing interaction.
The release matters to teams building customer service, guided walkthroughs and other face-to-face digital experiences. It is a product expansion beyond the Gemini 3.8 Live audio models introduced earlier in September, not a replacement announcement for those models. Google's evidence confirms availability and the intended capabilities, but it does not provide independent latency, quality or safety testing.
The delta is a continuing visual persona
Google describes Live Avatar as processing visual and audio inputs together and producing synchronized speech and video. The company says the generated persona can keep lip movement, expression and turn-taking aligned with the conversation. That changes the interface from a live voice session with multimodal input into a session where the model also maintains a visible speaking character.
The announcement says asynchronous tool calls can run while dialogue continues. In Google's example, an avatar can fetch information in the background without ending the active exchange. For builders, that is the more consequential systems claim: the visual presence is intended to persist while the agent performs work, rather than being a separate video response created after the task is complete.
Google also says the feature can move between 97 languages while adapting lip synchronization and expression. That is a vendor-reported capability, not a neutral measurement of language quality or visual stability. Teams considering multilingual use will still need to test accents, rapid code-switching, interruptions, noisy inputs and longer sessions in their own deployment conditions.
Enterprise availability has an important access boundary
Live Avatar is listed as available in Gemini Enterprise, and Google links to API documentation from the announcement. The company offers preset avatars, while custom avatar creation from a reference image is available only through enterprise allowlisting. The distinction is important: general Gemini Enterprise availability does not mean every customer automatically has the same customization rights.
The source does not specify public pricing for the avatar feature, regional availability, quotas, video resolution, measured end-to-end latency or the review criteria for custom-avatar access. Those missing details affect procurement and production planning. A product can be available while still requiring account-level approval for its most brand-specific capability.
Provenance is promised, but durability is untested here
Google says Live Avatar output carries SynthID in both audio and video. That gives customers a stated provenance mechanism for generated media. This run found no independent test of whether the watermark remains detectable after common edits, screen recording, recompression or deliberate removal, so the safeguard should be treated as a documented product claim rather than a proven guarantee.
The announcement also does not disclose an external evaluation of likeness preservation, misuse resistance or false attribution. Enterprises using custom identities will need governance beyond the model feature itself, including consent, access control, logging, review and clear disclosure to people interacting with an avatar. Those operational implications follow from the product boundary; they are not evidence that Google has solved every deployment risk.
What to watch next
The next verifiable checkpoints are the API's regional and quota documentation, broader access rules for custom avatars, published latency and quality measurements, and independent testing of SynthID persistence. Comparisons with other live-avatar systems should wait for matched tests rather than rely on separate vendor demonstrations.
For now, the supported conclusion is specific: Google has added generated visual presence to Gemini 3.8 Live and made the feature available through Gemini Enterprise, while some customization remains gated and its performance and safeguard claims remain uncorroborated.
Status
Confirmed. Internal confidence: Medium. One official Google announcement supports the release, availability and documented feature boundaries; no independent evaluation was fetched in this run.
Sources
Update note: Last reviewed 2026-09-25. We will revise this post if Google publishes broader access terms, measured performance or independent provenance testing.
Sources
- Google — Gemini 3.8 Live with Live Avatar — official
Drafted with AI assistance from source briefs; reviewed for citation completeness and label accuracy.