There are guides, courses, entire communities devoted to the art of the prompt. How to phrase the right instruction to get the best image. Which combination of words triggers which style, which light, which atmosphere. It is fascinating, and deeply incomplete. Because the prompt is not the beginning of the work. It is barely the middle.
Before typing anything, there is a series of decisions that have nothing to do with technique: what do we want to say? To whom? In what context will this image be seen? What emotion should it provoke, and why that one rather than another? What is the brand’s visual DNA, and how does this image fit into it? These are creative direction questions. And without them, the best prompt in the world only produces a beautiful image with no reason to exist.
But let us suppose all these questions have been answered. That we have a solid brief, a clear intention, and a well-built prompt. The generated image comes out. And that is where many people stop, and where the work really begins.
An AI-generated image is a sophisticated draft. A starting point, not a delivery. In our practice, between the first generation and the final image, there is retouching, colour correction, cropping, adjustments of detail that cannot be asked of an algorithm because they are matters of aesthetic judgment. There are also checks: does this image respect the client’s graphic standards? Does it work in different formats? Does it hold up in print as well as on screen? Is it technically usable: resolution, colour space, transparent background if needed?
AI has not abolished these steps. It has only given the illusion that they no longer exist. And that illusion is expensive, in wasted time, in disappointing results, in unhappy clients who thought that “generating an image” and “delivering an image” were the same thing.
In our process at artlequin, we use generative AI the way one uses a quick sketch: to explore directions, test atmospheres, validate an intention before investing in full production. It is a powerful prototyping tool. But the sketch is not the painting. And the painting demands craft, patience, and a precise knowledge of what you want to obtain, long before opening any software.
Deciding what to show, why, and to whom: that is the real work. Everything else (the tools, the techniques, the prompts) is only the means of getting there. This logic of prototyping as a decision tool, and not as an end in itself, is at the heart of what we describe in our article on why visualizing before producing saves projects.