Google Gemini App Launches Avatar Feature for AI Video Generation

AI Avatars Are Here: Google’s Gemini Now Lets You Star in Your Own AI Videos

Google has introduced avatars in its Gemini app, allowing paying subscribers to generate AI videos featuring a digital clone of themselves. This new feature is powered by the company’s Omni video model.

Available exclusively to paying subscribers on the Google AI Pro plan at $20/month.

To create an avatar, users simply read numbers and move their head in front of a phone camera—a process taking about five minutes. Once the avatar is set, users can generate videos by providing text prompts. The AI then places the avatar in specified scenes and performs requested actions.

Google restricts the feature to adult users creating videos only with their own avatar. This differs from OpenAI's previous approach, which allowed opt-out for others' use of likeness.

Example videos included a user singing to a dinosaur in Dolores Park, San Francisco, and surfing under the Golden Gate Bridge.

The feature is reminiscent of core capabilities in OpenAI's now-defunct Sora app. Users reported capacity limits: after two 10-second clips and a few questions, the system capped activity, with limits resetting every five hours.

• Background Fidelity: Generated backgrounds showed high fidelity to real locations, attributed to Google's mapping data.
• Visual Artifacts: Generated videos contained jumbled moments, nonsensical outfits, and inaccurate teeth.
• Appearance Capture: The setup process captures the user's current appearance, including clothing, which may appear in generated videos.