Google Gemini App Adds AI Avatar Feature for Video Generation

Google has added an avatar tool to its Gemini app that generates short videos of a user’s digital likeness. The feature is available only to subscribers of the company’s AI Pro plan, which costs $20 per month. A Wired reporter tested the tool by recording a short facial scan in a well-lit room.

The process took about five minutes and involved reading two-digit numbers while turning the head. Two ten-second clips were produced before usage limits reset after five hours.

One clip showed the avatar singing to a dinosaur at Dolores Park in San Francisco. The background matched the real location, including palm trees and the Salesforce tower, though the avatar’s clothing and movements contained errors. A second clip placed the avatar on a surfboard beneath the Golden Gate Bridge.

The avatar wore denim instead of a wetsuit and appeared in GoPro-style shots from the board.

DeepMind product lead Nicole Brichtova said the company aims to prevent harm while avoiding restrictions on benign uses. The system currently limits avatar creation to adult users generating videos of themselves. The reporter noted that the generated videos appeared more lifelike than older recordings of the same person, despite visible glitches in speech and clothing.