Whatsapp appears to be working on a new generative AI feature that should allow users to make personalized avatars of themselves for use in any imagined setting. The in-development feature, spotted in the new WhatsApp Beta for Android 2.24.14.7 by WABetaInfo, will seemingly use a combination of user-supplied images, text prompts, and Meta’s AI Llama model to generate the images.

A screenshot shared by WABetaInfo says that users can imagine themselves “in any setting from the forest to outer space.” The examples look fairly typical for those produced by AI generators, particularly if you’ve used apps like Lensa AI or Snapchat’s “Dreams” selfie feature.

To create the personalized avatar, WhatsApp users will need to “take photos of yourself once” which will then be used to train Meta AI to produce images in the user’s likeness in any setting. WhatsApp users will then be able to generate their avatars by typing “Imagine me” with a description of the setting in Meta AI chat, or in other WhatsApp conversations by typing “@Meta AI imagine me...”

The feature will reportedly be optional, and will require users to manually enable it in the WhatsApp settings before it can be used. WABetaInfo also says that the reference images can be deleted at any time via the Meta AI settings.

There’s no indication of when the new Imagine AI selfie feature will be generally available. WhatsApp is still rolling out support for its Meta AI chatbot, alongside more general real-time AI image generation for users in the US — so it may take a while before the new AI avatar feature launches. Meta would be right to take a more cautious approach given the issues seen with its previous generative AI tools.