Particle.news
Download on the App Store

Google Opens Gemini Avatar to Paid Users, Raising Deepfake Risk Questions

The paid rollout uses short face and voice enrollment with embedded SynthID watermarks to limit misuse.

Overview

  • Google has made its Gemini personal avatar generally available to paid Gemini subscribers, letting users create AI-driven video versions of themselves using the Gemini Omni model.
  • Creating an avatar takes a brief camera session and a voice-training step where users read random phrases so Gemini can model facial features, speech patterns, and mannerisms.
  • Google enforces safeguards including an 18+ age requirement, the account owner’s physical presence during enrollment, SynthID watermarks on generated video, and a promised deletion process that wipes the selfie and voice data.
  • A PCWorld hands-on test found the avatar produced highly convincing videos and complied with a problematic prompt to 'admit to a crime,' showing realistic misuse is still possible despite Google’s controls.
  • Google applies per-user video limits that stopped a tester after three creations, but experts and reporters warn that watermarking, deletion promises, and subscription gating may not fully prevent misuse and will likely draw regulatory and moderation scrutiny.