Gemini Shared Avatars: Images, Videos and the Limits of Access Controls

Google has announced shared personal avatars in Gemini, letting people include themselves and friends in generated images and videos. The feature combines reusable likenesses with sharing controls. The important limitation is that withdrawing access prevents new generations, rather than erasing content someone already created.

In its official announcement, Google describes tagging avatars in prompts and maintaining character and voice consistency. That is a product claim, not an independent guarantee that every output will preserve a person's appearance or voice correctly.

How Gemini avatars work

Google's avatar help page describes recording your face and voice, then selecting the avatar or mentioning your Google username with an @ sign. Images use Nano Banana; videos use Gemini Omni. Sharing is managed through Gmail addresses, and recipients can accept or decline.

The change is useful because a recurring character becomes something a collaborator can reference rather than rebuild in every prompt. For someone making a series of short scenes, that could reduce the repetitive work of explaining who should appear in each one.

It also changes the collaboration model. An avatar is an input that another person can reuse. Deciding who receives it deserves the same attention as deciding what the original prompt should say.

Who can use Gemini avatars

Google currently requires a personal account and an age of at least 18. The feature supports English and is unavailable in the EEA, Switzerland and the United Kingdom. Making avatar videos requires a Google AI plan.

Those restrictions matter before planning a collaborative project. A feature that works for one participant may not be available to another. The announcement does not establish a separate avatar API, new pricing or a universal rollout across every account type.

Removing access does not recall old images

Google says users can remove a recipient's access and manage expiration dates. Removing access stops that person making new avatar content; previously generated images and videos remain. Deleting the avatar also does not delete content already made and published.

Permission to create and control over a finished artifact are different things. A team might withdraw access after a project ends, but it still needs a separate understanding about where the resulting material can appear and how long it can stay there.

That distinction is easy to miss in a demonstration. The interface can make sharing look like a reversible switch, while the output becomes a separate piece of media. The sensible workflow is to agree on the intended use before sharing, then review the actual outputs before publication.

Consistency still needs review

For a creator, reusable avatars could make a recurring cast easier to manage. But consistency is only one part of a usable scene. A generated line can still say something the person would not endorse, and a familiar-looking face can make that line feel more authoritative than an obviously fictional character would.

A useful review therefore asks two questions: does the output preserve the intended likeness, and is the depicted action or statement appropriate for that person? Passing the first test does not answer the second.

The feature's value will depend on how well those creative and permission workflows fit together. Making a likeness easier to reuse is convenient; making its boundaries understandable is what allows that convenience to survive beyond a demo.

I’m Paras. I work at the intersection of AI, engineering and distribution.

I help technical founders turn product launches and complex AI features into clear stories that people can understand and share.

Book a 30-minute call.