Presented by
How to master AI prompts to create the image you need the first time
One of the biggest mistakes people make with AI image prompts is assuming the tool will somehow fill in the blanks for them.
In practice, Adobe Firefly works best when you treat prompting less like a vague request and more like a creative brief, with a clear subject and a defined look.
Firefly can work from style and composition references, while Photoshop now supports reference images for generative edits, and many other Adobe apps have new AI elements.
If you are building a quick social graphic, mocking up artwork for a feature, or creating something you plan to polish further in Photoshop, better prompting can a lot of save time.
We've compiled a few top tips and tricks to help you get the images you want.
Please note: All of the information is correct as of March 2026. Adobe regularly updates its products, so some steps or features may change.
Describe the image, not the action
Start with the picture itself.
In Firefly, prompts work best when they describe what should appear on screen, rather than sounding like an instruction to a tool.
Words such as “create”, “make”, or “generate” do not add much on their own, while a clear description of the image – the subject, the environment, the angle, and so on – gives the model something far more useful to work with.
Firefly’s current text-to-image tools are built around that kind of direct prompt writing, so a short, specific description will usually get you closer than a longer prompt padded with generic wording.
A simple way to approach this is to picture the finished image before you type.
If you want a product shot, say what the product is, where it sits, and how it should be lit; if you want an illustration, name the subject, the setting, and the visual style.
The wording does not need to be elaborate, it just needs to be visual, concrete, and deliberate.
Build every prompt around four essentials
One of the easiest ways to improve a prompt is to give it a simple structure before you start typing.
Break it down into four essentials: the subject, the setting, the style, and the mood. In practice, this means deciding what the image is about, where it takes place, what kind of finish you want, and how it should feel overall.
Take an AI-generated travel image. “Woman on a coastal train at sunset” gives you the subject and setting, but still leaves unknowns. Add “editorial photo” and “warm, reflective mood”, and the request becomes much more focused.
Using this technique also gives you a cleaner way to fix weak results. For example, if the image is broadly right but the atmosphere feels off, simply adjust the mood.
Use Firefly’s controls before rewriting
Don't assume the Firefly prompt box has to do all the work.
Firefly gives you a range of built-in controls that can shape the result before you start rewriting your wording, including content type, effects, colour, and lighting.
In practice, these settings can often fix the exact problem you are looking at far faster than typing another version of the same prompt.
If the image feels too flat, adjust the lighting. If the framing is wrong, change the camera angle. If the overall finish is off, switch the content type from Art to Photo, or explore the style options instead. And so on.
A concise prompt paired with the right settings usually gives you a cleaner result than a sprawling text request trying to force every detail into one line.
Use style references
Style references are one of the quickest ways to make Firefly's image output feel more predictable.
If you already know the look you want, a reference image can steer the output more effectively than trying to describe every visual detail in text.
Using a reference is especially useful when you are creating a set of assets that need to feel connected. A single campaign graphic might be easy to improvise, but consistency gets harder when you need several similar images.
The wording still matters, but the job becomes simpler. You can focus on the subject and scene, then let the reference image anchor the broader look.
Use composition references
Sometimes, the issue is not the look of the image, but the layout.
If you already know where the subject should sit, how the frame should be balanced, or how much space needs to be left around key elements, a composition reference gives Firefly a much clearer starting point.
Upload a source image, and the tool can use its outline and depth as a guide for the new result, with a strength slider to decide how closely it should follow the original structure.
For hero images, product shots, or any visual where positioning matters, that can be far more reliable than repeatedly nudging the prompt and hoping the framing lands in the right place.
Photoshop now supports the same broader approach through Generate Image, including reference images for composition, so it is easier to carry a layout from early concept to later edits without rebuilding it from scratch.
Know when to stop prompting
At a certain point, rewriting the prompt stops being the fastest route.
Once the subject and general look are in place, it often makes more sense to switch into editing and shape the image with intent.
Photoshop is built for that handover. Generate Image can use a reference image to guide either style or composition, which helps you keep the direction you have already established while you explore variations.
If you are working towards something publishable, that combination of a solid first prompt alongside targeted edits is usually quicker than chasing a perfect one-shot result.
