Meta – Avatar Creation Reimagining how a billion users create their digital self, from stylised to photo-realistic avatars Context I led design across Meta’s avatar creation platform for 3 years – spanning several interconnected initiatives Avatars are a core part of how people represent themselves digitally, and until recently, creating one meant manually selecting every facial feature from scratch. We were among the first in the industry to allow users to generate an avatar from a selfie or photograph – and then pushed further, enabling photo-realistic avatars through a more robust face scan, to improve co-presence and remote communication. Most users who tried to create an avatar gave up, or created one they never actually used. “Every time I try to make an avatar that actually looks like me, I end up spending ages tweaking sliders and options – and after all that effort, it still doesn’t look anything like me. Why bother?” – Meta Avatar user Problem The tools asked too much of users and gave too little back: Creation friction: With hundreds of assets to browse across every facial attribute, users struggled to identify options that matched their own features High abandonment: The sheer number of steps meant most users dropped off before ever reaching a finished avatar Poor likeness: Users who completed the flow were often dissatisfied – manual selection rarely produced a result that resembled them Styling: Keeping the avatar dressed and styled required the same manual browsing every time users wanted to change their look Solution We replaced manual creation with smarter, image-based flows – giving users the confidence and control to get to an avatar they’d actually use. Photo-based creation: Avatar generation from a live selfie, profile picture, or device folder – faster, more accurate, and far less effort Quick Edits: Lightweight post-generation adjustments to tweak or correct key facial features without starting from scratch Photo-real avatars: A new robust face-scanning flow – the first of its kind – enabled photo-real avatars for improved co-presence AI styling: I set the direction for an AI-powered editing flow where users could simply describe what look they wanted Outcome New creation flows drove a step-change in completion and engagement at a scale of over a billion avatars. 5x increase in avatar creation 3x increase in DAU 35% increase in creation completion rate Patent application filed for Avatar creation from photos; with design patterns becoming foundational to all creation paradigms