WhatsApp is apparently developing an AI-powered image maker that will allow users to create AI avatars of themselves to place in virtual environments that represent them. The feature, first identified by WABetaInfo in a future beta release of WhatsApp’s Android version, will likely leverage WhatsApp parent company Meta’s Llama portfolio of AI models.
“This feature lets you take photos of yourself once and then ask Meta AI to generate AI images of you,” the beta screenshot explains. “To generate an AI image of yourself, type ‘Imagine me…’ in your Meta AI chat. You can also use this feature in other chats by typing ‘@Meta AI imagine me…’
Users would upload some photos of themselves and Meta’s AI Llama model would create an AI avatar of them that could then be placed into any image setting based on text prompts. While the feature is currently in beta testing and available to a limited group of users, it’s unclear when it will be widely available.
Avatars aren’t WhatsApp’s first AI feature. The social media platform recently added custom in-app stickers that users can create using text prompts. But making it possible to embed yourself in an AI-generated image would be a big step forward for the company.
AI Identity
The idea alone makes a lot of sense as a compelling feature for WhatsApp. But it would also stand out among Meta’s peers. AI image generators from companies like OpenAI and Google have typically been reluctant to create AI avatars of a real person, let alone one that could be reused repeatedly. ChatGPT almost consistently states that it can’t create an image of a real person, and Google Gemini resists the idea of replicating a person in favor of an image of a character with a similar likeness or simply wearing similar clothing. Google even recently made it easier to remove unauthorized AI avatars on YouTube, instituting a process for those who spot deepfake versions of themselves to request their removal.
Privacy and security are obvious concerns surrounding AI avatars of real people. While WhatsApp hasn’t yet detailed the specific privacy measures for the AI-generated avatars, it will likely have to have some strict rules in place when the feature launches. The company wants to ensure that user data is handled securely and that users have control over how their likenesses are used, and making the feature optional is a step in that direction. That the feature appears to use the Meta AI interface now available on Facebook, Instagram and other Meta properties suggests that WhatsApp could be a testing ground for an eventual broader rollout.