How it works today
- Searching stock photos everyone has already seen — 45 min
- Planning a shoot that isn't worth it for a single image
- Publishing posts without an image because time ran out
- The visual language changes with every post, because there isn't one
How it works afterwards
- A fixed visual language defined — light, colours, framing, mood
- A new image in minutes that recognisably belongs to the brand
- Motion where a still used to be
What we actually build
1. Define your visual language
Before any image is created, we settle how your images should look. Not "high-quality" — concretely: which light, which colour world, which focal length, which mood? That's design work, not tool work, and it decides everything that follows.
2. Learn the language of description
The visual language becomes reusable description modules. Along the way you learn the terms that actually steer image impact — lens type, lighting, camera movement, film stock. That's the part almost everyone skips, and it makes the entire difference.
3. Consistency across many images
One good image is luck; twenty images that belong together are craft. We build the path to recognisable visuals — same person, same environment, same feel across an entire campaign.
4. Motion, when it's worth it
Short video sequences from stills, animated logos, product movement. The effort is higher than for stills, so we do this deliberately where movement actually adds something — not everywhere.
Tools — and where it doesn't help
For stills there are several good routes with different strengths: some keep people consistent across images, others are stronger for graphic work. For motion the field is smaller and results vary more — rejects are part of the process, and anyone who doesn't budget for that ends up disappointed.
Where it doesn't help: anything that has to be real. Your team, your shop, your product in a real person's hand — that takes a camera. And an AI image faking a real situation becomes a liability the moment someone asks.
With images, it's about personality rights
As soon as recognisable people are involved, the right to one's own image applies — even when the image is generated and merely resembles the person. Uploading a photo of an employee as source material means processing biometric data. And on client projects, the question arises whether generated images are contractually permitted at all. We settle these points before production.
→ How we bring law and AI together
What this means in time
For a fitting image instead of a stock-photo hunt: from 45 down to 10 minutes. The bigger effect lies elsewhere — images that simply didn't exist before, because a shoot was never an option for them. That's not a time saving; that's a possibility space that didn't exist.
Frequently asked questions
Don't AI images instantly look like AI?
The typical tells — waxy skin, overblown light, six fingers, symmetrical composition — come from vague descriptions. With precise specifications for lens, lighting and film stock they largely disappear. Exactly that precision is what the sessions teach.
May I use AI images commercially?
With most major providers on paid tiers, yes. But there are exceptions and traps: trademark and personality rights apply regardless of the tool, and generating an image that resembles a protected character or a real person is a problem — even if the provider permits the use.
Do I have to label AI images?
Legally in Switzerland currently only in specific cases; in the EU increasingly strictly. My recommendation goes beyond the obligation: label anything that depicts a real situation. The loss of trust when someone else discovers it outweighs the cost of a note.
How long until I get usable results?
First usable images in the first session. A consistent visual language across an entire campaign takes three to four sessions — that's the difference between one good image and a system that delivers good images.
Bring three images you like
In the free intro call we take them apart — and you'll see how precise image description actually works.