- kind
- provider-policy
- target
- voice-and-likeness
- jurisdiction
- Google (industry)
- mechanism
- Lets a Google Account holder create a personal AI avatar of their own face and voice for use in Gemini Omni video generation inside Google Vids, gated by an on-device capture and verification step, an 18-and-over age requirement, and a domain-level admin on/off control; every generated clip carries an invisible SynthID watermark. Personal avatars are restricted to the account holder's own likeness, and the feature is withheld entirely from the EEA, Switzerland, and the UK.
- status
- in-effect
- owns
- Distinct from heygen-avatar-consent-tiers and synthesia-custom-avatar-consent, which both build a third-party consent workflow (webcam statement, pre-recorded video, or waiver) so a customer can clone someone else's voice or likeness. Google's policy is affirmatively scoped to the account holder's own likeness only, and it geofences the feature out of the EEA/UK/Switzerland rather than relying on a uniform worldwide consent standard.
- notes
- Feature announced July 16, 2026 (Rapid Release start) with the Scheduled Release track, covering the bulk of Workspace domains, beginning August 5, 2026, the first day of this lens's search window. The self-only restriction is confirmed verbatim on the companion official blog post (blog.google/products-and-platforms/products/workspace/gemini-omni-personal-avatars/): "Personal avatars are linked to a user's Google Account and are restricted to the account holder's likeness." Companion support page (support.google.com/gemini/answer/16984474) states verbatim: 'Be 18 or over', 'Be signed in with a personal Google Account', 'Avatars are currently not available in the EEA, Switzerland, or the United Kingdom', and 'The account owner must complete this process. Do not allow minors to record themselves.' Status normalized to 'in-effect' per this task's schema enum (in-effect / proposed); note the five existing provider-policy rows in the baseline instead use 'active', a divergence the orchestrator may want to reconcile across the dataset.