How it started

My name is Nils, and I’m an industrial and experience designer. Pretty Normal People started with a frustration I first encountered while working on my master’s thesis. I was developing a wearable device designed to help healthcare professionals support older people in managing their hydration. Part of the project involved a small patch worn directly on the skin, and I wanted to show what it would look like on the people I was actually designing it for. What sounded like a simple task turned into weeks of searching.

I went through stock photography websites, Pinterest and Google, trying to find photographs of older people’s skin that I could realistically Photoshop my design onto. I needed the right perspective, lighting, skin texture and enough resolution to make the result convincing. But most of what I found was either unsuitable or looked nothing like the people I had met during my research. Generative AI wasn’t nearly as accessible or capable back then, so my options were limited. I remember thinking how strange it was that I could spend months researching people’s needs and developing a product specifically for them, yet struggle so much to simply show that product on a representative body.

And I don’t think I’m alone in this. As designers, we spend a lot of time understanding people. We speak to users, study their behaviour, observe their routines and try to understand what makes their lives different. But when it comes to visualising our ideas, we often end up relying on whatever imagery we can find. Of course, the ideal solution would be to return to the people we designed for, have them interact with our prototypes and photograph everything professionally. But anyone who has worked on a student project, a tight deadline or a limited budget knows how difficult that can be.

A few years later, I began experimenting with generative image tools in my design work. Suddenly, something that had previously taken hours of searching and compositing seemed possible within seconds. You could place almost any product on almost anyone, in almost any situation. At least in theory. In practice, I kept running into another problem. The people appearing in my images looked remarkably similar. Beautiful, young, symmetrical, often impossibly smooth-skinned. They looked more like advertising models than the people I encountered in everyday life. Even when I specifically asked for something different, the results had a strange tendency to return to a very narrow idea of what a person should look like.

I found that fascinating. The more I looked into it, the more I began to understand why. Generative AI learns patterns from enormous amounts of existing imagery. And much of the imagery we produce and publish, particularly in advertising, fashion and commercial photography, represents a fairly limited range of bodies, ages and appearances. These patterns inevitably influence what the tools generate, even when we ask them to do otherwise.

What began as a series of small experiments slowly developed into a much larger personal research project. I started exploring how to create characters with believable anatomy, natural skin, individual proportions and all those seemingly insignificant details that make someone recognisable and real. It turns out that generating an ordinary-looking person is surprisingly difficult.

Creating one convincing image is one thing. Making that same person recognisable from different angles, with consistent skin, proportions and features, is something else entirely. Much of my work now involves exploring exactly these challenges, testing different tools, developing workflows and trying to understand where generative AI works, where it struggles and how much creative judgement it still requires.

Pretty Normal People is my way of sharing what comes out of that research. The Instagram page and character library are collections of people developed through these experiments, made available for other designers, makers and creatives to use in their own work. In a way, I’m trying to create the resource I wish I’d had when I was working on my thesis.

Of course, none of these characters are real people. They are artificially created, and they cannot replace the presence, experiences or individuality of someone photographed in real life. Nor should they replace the people we meet and learn from during design research. But I do think there’s something worth exploring here. Especially in those early moments of the creative process when we want to imagine a product, an experience or an idea with someone who doesn’t necessarily fit the usual visual conventions.

I’m still learning, experimenting and questioning much of what these tools can produce. That’s really the point of this project.

Because when we spend so much time designing for the diversity of everyday life, shouldn’t we be able to show it, too?