How I work.
People have been asking how I make the images for years. Lately they ask about the AI.
I want to try to explain it for real. Not to defend anything — but if you're going to buy one of these pieces, you deserve to know exactly what it's made of.
Before AI.
I started drawing because it was the one place I felt safe. I was twelve. Getting along with people was hard, and the paper held everything I couldn't say out loud.
Then came painting. Then photography. Then photomontage — putting things together that had no business being together and seeing what came out. Then digital illustration. This was the early 2000s, well before anything like what we now call generative AI existed. I worked with my own photos, archive images, pieces of things I cut out, layer after layer in Photoshop. The result wasn't a paper collage: it was a new image built from the inside.
Zoo Portraits started in 2012 the same way. A fox. A suit. A neutral background. No theory behind it. And it connected — which is still the part I don't fully understand.
When generative AI arrived.
When Midjourney and tools like it became actually usable, I didn't ignore them. I didn't adopt them all at once either. I tested them for months, set them aside, came back. What interested me was what they are: generators of raw visual material that didn't exist before.
What they do in my process is specific. I ask for fragments: a pose, a coat texture, a lighting angle I can sense but don't have. I give them references from my own style, tell them I want something that sounds like what I already do. The model gives me something back. Most of the time it's not what I want. When there's a fragment I can use, I take it into Photoshop.
That's when the familiar work begins.
What happens in Photoshop.
The AI-generated fragment isn't the piece. It's a starting point — like an archive photo or a pencil sketch used to be. Sometimes less clear than that.
What happens next is photomontage, and it's how I've always worked. I build the animal by combining several sources: licensed animal stock photography, AI-generated fragments, digital illustration, and occasionally my own photos for texture. Before AI, that collage was stock and illustration; now I add generated fragments to the mix. I put it all together by hand in Photoshop, layer by layer: the symmetry, the light, the color, the shadows, the way the parts come together. I retouch. The character gets assembled slowly — sometimes over days — with a manual decision at every step.
I use Midjourney and Google Gemini to generate those fragments. I don't train any model myself. I use general-purpose models and ask them, literally, to imitate the style of "Yago Partal" and "Zoo Portraits." And there's an irony I'm not going to stay quiet about: they manage it — the animals come out symmetrical, frontal, dressed — almost certainly because my work has been circulating for over a decade and the models have absorbed it. The same tool that helps me is, in part, copying me back at myself.
What you receive is the result of that manual work. Not the AI output. What I built on top of it.
Why I don't call it something else.
There are cleaner-sounding ways to describe it. "Digital art," full stop. "Digital photography." "Illustration." Several of those are half-true and leave something important out.
I call it photomontage because that's what best describes the operation: combining fragments from different sources, compositing by hand, making the whole into something none of the parts were on their own. The fragments I use now include — alongside stock and my own images — material generated with AI. That's new. I'm not hiding it.
What I'm also not going to do is call it "AI art," because that label describes something different: images where the model generates the final result and the artist selects from options. That's not what's happening here. What's happening is that the AI generates intermediate material, and I do the work I've always done.
One point, to avoid any confusion: this applies to my digital work and to Animal Kinhood. When I make pure photography or pure illustration, there's no AI in it. Those are different things and I don't mix them.
What this means for what you're buying.
Every Animal Kinhood edition carries a technical record embedded in the digital file, in a format called C2PA, that documents exactly which tools were present in the process. It isn't hidden. You can read it.
The technical descriptor we use is compositeWithTrainedAlgorithmicMedia. That's not "AI-generated art." It's an authored composition where part of the source material comes from generative models. There's a difference, and the international standard recognizes it.
What I sign is the resulting piece. What I guarantee is that every compositional decision — light, color, integration, retouching — was made by me. The AI didn't make those decisions.
A question I sit with.
Where's the line between using AI as a tool and letting AI make the work?
I don't have a clean philosophical answer. What I do know is that in my case, if I strip out everything that isn't AI, I'm left with a blurry, characterless fragment that wouldn't mean anything to anyone. What makes a character recognizable — the look, the sense of a life behind it — all of that comes from the work that follows.
And if you ask whether I think that's enough: yes. But I understand it might not seem that way to other people, and I think it's fine for them to think so.
Where the process stands now.
Today, in 2026, the workflow is what I described: stock, fragments from Midjourney and Google Gemini, illustration, and manual photomontage in Photoshop as the main process. If I ever train a model on my own archive — I've thought about it, I'm not ruling it out — I'd have more control over the source material. If that day comes, I'll say so.
What won't change is the compositional method. Or the manual work that gives each character its shape. Or the decision to document the process with the same honesty as what's written here.