You can look at a photograph and know that you want it to feel different without knowing how to explain the difference. Perhaps the room looks too clinical, the colours seem too loud, or the light feels wrong for the scene. “Make it warmer” is a start, but warmer could mean afternoon sunlight, softer shadows, more yellow in the palette, or a sense that someone has just stepped out of the frame.

That uncertainty offers a useful starting point for a creative exercise. Take one ordinary photograph, find a few references that interest you, and explore what happens when you translate your preferences into specific instructions.

The aim is to finish with a clearer understanding of the choices that make an image feel right to you.

Consider a photograph of a blue cup on a wooden kitchen table, with a small plant beside the window. It is a modest scene, but it gives you plenty to work with: the curve of the handle, the grain of the wood, the colour of the leaves, and the way daylight falls across the surface. For this exercise, those familiar details provide something to return to as the image changes.

Before generating anything, write down what you already like. Perhaps it is the space around the cup or the shadow beneath its handle. These details belong in the brief just as much as the things you want to change.

An AI prompt library can help with the next step. Promptsref lets readers browse generated work and examine the associated prompts, with categories covering different models and visual styles. Looking at an image alongside its description gives you material to study when your own vocabulary feels imprecise.

Choose just three references. One might interest you because of its muted colours, another because of its lighting, and a third because of its illustrated texture. For each, complete a short sentence: “The part I want to try is…”

This requires a closer look than simply saving an attractive image. You might discover that a scene you initially described as “cosy” contains fairly cool colours. What appeals to you is the small pool of light around the subject. Another image might feel nostalgic because of its faded contrast rather than anything depicted in it.

Bring those observations back to the kitchen photograph. Leave behind instructions that belong to the reference’s subject or setting. A description written for a crowded street may contain useful lighting language, but the crowd, architecture, and camera position have little to contribute to a picture of a cup.

The first experiment could focus entirely on light.

Promptsref’s Image to Image AI Generator accepts a reference image and written instructions describing the intended changes. For the kitchen scene, an example brief would be: “Keep the blue cup, plant, table, and framing. Use soft late-afternoon light from the window, with gentle shadows and visible detail in the wood.”

Treat that instruction as a request to evaluate. The resulting image still needs inspection, particularly where the cup’s shape, the plant, and the arrangement of objects matter.

Place the variation beside the original and ask a specific question: does the new light make the table more inviting, or does it simply make everything orange? If the colour cast is too strong, the next instruction might ask for more neutral whites while retaining the softer shadows.

That revision expresses a more precise preference than “make it warmer”.

For a second experiment, return to the original photograph and explore colour. Ask for the blue cup to become a muted terracotta shade while retaining the existing lighting and surrounding objects. Starting again from the original gives you a clearer basis for comparison than editing a version whose light has already changed.

Notice where your attention goes. Does the cup stand out more strongly against the leaves? Does the new colour compete with the table? Either response can inform the next decision. You may decide that the original blue was doing useful work.

A third experiment could explore an illustrated treatment, using restrained coloured-pencil marks and areas of visible paper. Here, the question becomes how much detail the scene can lose while keeping its character.

If the wood grain becomes distracting, simplify it. If the handle loses its familiar shape, make that shape a priority in the next brief. An attractive treatment can still be wrong for the particular image you want to make.

Throughout the exercise, keep a brief note beside each version: what you requested, what visibly changed, and what you would keep. A changed prompt does not guarantee that only one feature will change, so describe what you can actually see rather than assuming the instruction worked exactly as intended.

Stop after a small set of variations and choose one detail worth carrying into another project. It might be a preference for softer shadows, a dislike of heavy texture, or an appreciation for the blue you nearly removed.

The next time you open a photograph, you will have something more useful to say than “make it better”. You might ask for gentler light while keeping the cool colours, or a simpler surface without losing the object’s shape. Those decisions give the next image a direction—and they are decisions you can recognise as your own.

Amelia Hart, a psychology graduate from the University of Hertfordshire, has a keen interest in the fields of mental health, wellness, and lifestyle.