Gemini’s AI avatar tool creates realistic personal video clones from a short phone setup

Google’s Gemini app can now generate short AI videos using a digital recreation of a subscriber’s face and voice, moving consumer AI video closer to fast personal likeness generation.

The avatar feature is powered by Google’s Omni video model and is available to Gemini subscribers. Setup takes about five minutes in the Gemini app. The user records their face, reads a series of two-digit numbers, and turns their head so Gemini can build a personal avatar.

A hands-on WIRED test produced two 10-second clips: one showing the reporter singing to a dinosaur in San Francisco’s Dolores Park, and another showing the same avatar surfing under the Golden Gate Bridge. The videos included errors, odd clothing, and chaotic object behavior, but the face and voice likeness were close enough to feel unsettling.

Google currently limits adult users to generating videos with their own avatar. Google DeepMind product lead Nicole Brichtova described the company’s safety goal as preventing harm without blocking benign uses.

The feature arrives as AI video tools continue to raise questions about consent, identity misuse, and nonconsensual deepfakes. Gemini’s implementation narrows the feature to self-use, while still showing how quickly high-fidelity personal video generation is entering mainstream consumer apps.

Source: WIRED, “I Cloned Myself With Gemini’s AI Avatar Tool. The Result Was Unnervingly Me,” May 21, 2026 — https://www.wired.com/story/i-cloned-myself-with-geminis-ai-avatar-tool-the-result-was-unnervingly-me/