Google is expanding access to one of Gemini's most unusual features. The company is now rolling out Gemini Avatar more broadly to paid subscribers in the Gemini app, allowing users to create AI-generated versions of themselves using their face and voice.
The feature, which was first identified earlier this year in an APK teardown, creates a digital clone that can speak in video clips generated through Gemini. Google says the tool is powered by its Omni model, the same system used to reconstruct a user's appearance and voice for avatar creation.
Avatar setup is available inside the Gemini app through Settings > Avatar. To build an avatar, users go through a guided enrollment process that requires them to look into the camera, move their head, and read out numbers. Google uses that recording to map facial details and capture voice characteristics.
Once the setup is complete, users can bring the avatar into Gemini chats with simple prompts, including commands such as @me or by using their account name. The result is an AI version of the user that can appear in generated videos while closely matching their look and speech patterns.
The company has described the feature as a way to add yourself directly to video creations. In practice, it produces a highly realistic digital stand-in rather than a stylized cartoon version, which may make the output feel more convincing, and more unsettling, to some users.
Because a tool that can generate a believable copy of a person's face and voice could be misused, Google is putting limits around access. Avatar creation is restricted to users who are at least 18 years old, and the account holder must be physically present during the setup process.
Google is also adding a watermarking measure to the content produced by the feature. Every video made with the Omni-based avatar system includes an invisible SynthID watermark embedded in the file. Google says that watermark can be checked through Chrome or Google's Search tools to verify that a clip was generated by AI.
The wider rollout comes months after the feature first surfaced in testing. Google has not framed Avatar as a general public feature yet, but the broader release to paying Gemini users suggests the company is moving carefully toward a more public launch.
For now, Avatar joins a growing set of generative AI tools aimed at making video creation more personal. It also raises the same questions that have followed other synthetic media features, including how realistic digital likenesses should be labeled, who should be allowed to use them, and how platforms can prevent misuse.
Google's latest expansion shows that it is still pushing forward with highly personalized AI features, even as those tools continue to test the boundaries between convenience, creativity, and impersonation.