Visualising ideas with Dall-E

Closing the gap between language and image

The most useful thing GenAI has given me isn’t “art”, and it isn’t speed for its own sake. It’s something far simpler and, if you spend your life in ideas, quietly liberating.

It has shortened the distance between a story in your head and an image you can point at.

If you work in communications, brand or experience, you’ll recognise the gap immediately. Ideas tend to sit in language for far too long; we describe a scene, a mood, a world, everyone nods, everyone imagines something slightly different, and then, weeks later, someone “makes it visual”, at which point the conversation finally becomes real and everyone discovers what they were actually disagreeing about.

GenAI doesn’t remove that mess. It lets you enter it earlier. It gives you a way to rough out a visual direction fast enough to think with it, change your mind, test alternatives and reach alignment before the project calcifies into decks, deadlines and polite misunderstanding. This is what I mean when I talk about the augmented studio, not as a grand concept but as a daily shift in how you work; you stop waiting for “the visual phase” to arrive later, and you start exploring visual thinking alongside the strategy, while the idea is still light enough to move.

That’s the crux for me: visualisation as a creative thinking tool. Not as finished output, not as replacement for designers, but as a way to pull an idea out of abstraction and place a version of it on the table while it’s still malleable.

The loop that changes everything

The workflow is almost boring, which is precisely why it’s powerful.

You write a brief as if you’re speaking to a top-tier designer. You include constraints. You describe tone, materials, proportion, light, emotional temperature. You generate. You judge. You refine. You repeat. ChatGPT becomes the thinking partner that helps structure the brief and sharpen variation without losing the plot, and DALL·E becomes the visual engine that turns those instructions into something you can react to.

The iteration speed is the real shift. Not one direction. Not “the answer”. Multiple plausible interpretations in the time it once took to schedule a first round. That unlocks something that used to be reserved for those who can sketch, or for teams with budget to explore; you can now audition visual directions the way you audition words, and you can do it early enough that it genuinely improves the thinking rather than merely decorating it.

But acceleration comes with a useful constraint: when the machine can generate endlessly, taste becomes the differentiator. The tool will always give you something, often shiny, sometimes ridiculous, occasionally genuinely good, but it cannot reliably tell you what’s true, what’s on-brand, what’s credible or what should be left out. That filter is still human, which is why I keep returning to taste as the real skill, not prompting.

Rebuilding what’s missing

I started testing this in small ways, partly because that’s how you learn where the edges are.

Neu-Falkenstein in the Jura has become a regular family walk since lockdown; the kids climb the stones, ask questions, and my brain does its usual thing, quietly rebuilding rooflines and courtyards from whatever the weather has left behind. Using GenAI here isn’t about claiming accuracy. It’s about turning that vague mental reconstruction into something you can see, critique and adjust.

DALL·E will absolutely try to romanticise everything. It tidies up the stonework, adds improbable drama, gives you skies that suggest a Netflix budget. You have to rein it in with constraints, repeat what you mean and accept that you’re producing a visual hypothesis, not a historical document.

When it starts to click, the value is surprisingly grounded. You feel the scale again, the approach, the logic of the ridge. The story becomes easier to tell because you’re no longer relying on hand-waving.

The same instinct applies in Basel. The surviving medieval gates hit hard precisely because most of the walls are gone. Rebuilding the approach to Spalentor as a lived scene, dusty road, cart tracks, travellers approaching, restores context. You stop seeing the gate as a static object and start seeing it as part of a journey. That’s often what our work is about anyway: restoring context so meaning makes sense.

Inventing what doesn’t exist

The opposite test is invention.

After a trip to Rulantica, the kids decided what they wanted next was Fasnacht. No theme park exists, which made the idea more appealing rather than less. Building a fictional indoor “Fasnacht World” became an experiment in experience thinking. What does this cultural idea look like as a place. Where does the eye land first. When does detail become unreadable. When does spectacle tip into nonsense.

I pushed some outputs into a pencil-sketch restraint at the end, not because I was pretending to be an illustrator, but because I wanted to see whether the idea survived without polish. If a concept only works when it’s covered in visual fireworks, it isn’t a concept. It’s decoration.

Again, taste quietly does the heavy lifting.

A tradeshow booth as proof

Eventually the experiments moved closer to home.

I wrote a master brief for a fictional “alotté pianos” tradeshow booth, including dimensions, materials and a defined visitor journey. ChatGPT structured the variations. DALL·E generated multiple layout directions in a single sitting. There were glitches, floating elements, odd shadows and physics-defying proportions, of course, but the speed of exploration was hard to ignore.

Within an hour I had genuinely different spatial directions I could talk through, not just gesture at. The augmented studio isn’t you becoming a production designer overnight. It’s you becoming more visually articulate earlier, so that when the real design team enters, the conversation starts from a sharper place.

This is acceleration, not delivery.

Building a house style

The same loop shaped the visual identity of this site.

I always knew I wanted a distinct house style, something you could recognise in seconds without needing a logo, something that delivered that quiet internal moment of yes, that’s exactly my taste. The early foundations were laid years ago inside Episerver at Bain & Company, shaping careers pages and learning how structure sets emotional temperature. It wasn’t glamorous work, but it taught me that spacing, rhythm and restraint are choreography, not decoration.

When alotté began to form, that discipline returned almost instinctively: minimalism, earthy tones, generous space, nothing forced. Two visual lanes emerged that feel like siblings rather than strangers. One sits in a restrained editorial illustration style, thin ink linework, visible pencil texture, muted palette and calm observational figures. The other leans into warmer, dreamlike photorealism, slightly cinematic but grounded in the same restrained colour world.

Both were built through iteration inside DALL·E and Gemini, not by accident but by refinement. Hundreds of tiny adjustments. Adding detail, removing detail, widening the frame, muting the palette, returning everything to that earthy, spacious foundation that alotté sits on. After a while the images stopped feeling generated and started feeling authored.

That’s the quiet point underneath all of this.

The real shift

GenAI rewards constraints, clarity and point of view. It punishes vague briefs and lazy thinking. It gives you more options than you need, which forces you to become sharper about why you’re choosing one over another.

If you can describe the story clearly, you can see it sooner. If you can see it sooner, you can shape it sooner. If you can shape it sooner, you waste less time pretending everyone imagined the same thing when they didn’t.

The augmented studio, in this context, isn’t a futuristic concept. It’s the practice of bringing visual exploration forward into the moment where it can still improve the thinking.

And taste, not technology, is what makes the output coherent, credible and worth anyone’s attention.