There are moments when a creator decides to abandon an approach that worked - not because it stopped working, but because they found something more interesting. That's what happened to Barak Rotem, AI artist and visual storyteller, when he decided to stop chasing the perfect prompt.
"I stopped trying to write better prompts," he writes. "Instead, I started describing materials, texture, and structure more clearly." And from there, everything changed.
Luma's Uni-1 - A Model That Understands Materials
His experiment was made with Uni-1, a Luma AI model, which allowed him to create a small editorial series where familiar subjects shift into lace. Not lace as decoration. Not as an effect. But as a genuine material transformation.
The results Rotem published are impressive not just technically. They carry a quality that's difficult to explain: they look as though part of the subject was always made of lace, as if the photograph revealed a structure that was there from the beginning. This is one of the most interesting ways to think about AI visual expression - not as "what the model can generate" but as "what was already there, only now revealed."
The Point About Prompts
What Rotem subtly points out is a deeper shift in approach. Most AI visual users spend a lot of time searching for the right phrasing - what to write, which language to use, which keyword will produce the desired effect. That's logical, but it also has a limitation: it's phrasing, not description.
When you move from "write a good prompt" to "describe the material you want to see" - something different happens. You're not directing the model to arrive at a result, you're telling it what's in the image you see in your mind. That's a cognitive difference, not just a linguistic one.
Rotem - who comes from a visual art background rather than a technical one - arrives at this insight perhaps naturally. An artist accustomed to thinking in materials, textures and light translates that thinking into language the model understands better.
The Series as Manifesto
The series he created isn't just a technical demonstration. It poses a visual question: what happens when a material we recognize as soft and fragile - lace - becomes a structural element? When faces aren't decorated with lace but made from it?
This is the question art continuously explores: what happens when the boundaries of material dissolve? Rotem finds in today's AI tools a space to explore such questions in ways that weren't possible before.
The Biennale, Bezalel, and Now Luma
Rotem recently published about his participation in the UAE Biennale and the lecture he gave at Bezalel Academy. He's a creator who moves between the art and academic worlds, bringing the same seriousness and sensitivity to everything he publishes.
The new series with Uni-1 is a direct continuation of that trajectory: not using AI because it's a trend, but because the tools allow him to ask questions he couldn't ask before.
And that, ultimately, is the reason to follow what Rotem does - not because of the technical performance, but because of the questions he asks.