How to Create Marketing Images With AI: Prompts and Edits
A repeatable three-step routine for turning a plain idea into a usable marketing image, then keeping every image after it consistent with the first.
Eight AI Image Styles Worth Learning, With Prompts That Work
'Make it look cinematic' is not a prompt. Here are eight styles broken down into the specific words that produce them, and what to change when they miss.
ChatUp GuidesFun
We are well past the era of smudged hands and melted faces. Modern image models will do more or less what you ask — which is precisely why the limiting factor is now the asking. “Make this look like a movie poster” produces a movie poster in the same way that “make me dinner” produces dinner: technically yes, but not the one you wanted.
What follows is eight styles, each broken into the components that actually control the result. Read them as recipes: the words are doing specific jobs, and you can swap ingredients once you know which is which.
Part 1: Stylized
Cartoon movie poster
The trick is naming the era and the printing process, not just “cartoon.”
Prompt to try
A 1960s hand-painted adventure film poster. Bold flat color blocks, limited five-color palette in ochre, teal, cream, black and red. Slight offset print misregistration and visible paper grain. Dynamic diagonal composition, heroic low angle, dramatic scale contrast between the figure and the landscape. Large empty band across the bottom third for a title. No text.
The controls: era (“1960s hand-painted”), production method (“offset misregistration, paper grain”), palette limits, and composition geometry. Change the era first when it misses — it moves the result more than any other word.
Studio anime illustration
Prompt to try
Hand-painted anime background art in the style of late-90s theatrical animation. A quiet residential street at dusk, overhead cables, vending machine glow, wet tarmac reflecting warm light. Painterly cel shading, soft gradient sky from amber to deep blue, visible brushwork on the clouds. No characters, no text.
The controls: “background art” rather than “anime” gets you the painted environment rather than a character portrait. “Cel shading” and “visible brushwork” keep it illustrative instead of drifting photographic.
Part 2: Hyperrealistic
Editorial fashion cover
Prompt to try
High-end fashion editorial portrait, medium format film aesthetic. Single subject, three-quarter turn, direct eye contact. One large softbox camera-left with a subtle rim light behind. Matte skin texture retained, no beauty retouching, natural pores visible. Seamless muted sage backdrop. Shot at f/4, 85mm equivalent, shallow but not extreme depth of field. Cool neutral color grade. Vertical, generous headroom for a masthead.
The controls: the lighting setup and the lens do nearly all the work. “Matte skin texture retained, no beauty retouching” is what prevents the plastic look that marks an image as generated at a glance.
Candid street photograph
Prompt to try
Candid documentary street photograph, 35mm, available light only. Rainy weekday morning, a person mid-stride passing a shuttered shopfront, umbrella partially obscuring their face. Slightly imperfect framing, one element clipped by the frame edge. Natural motion blur in the background, grain visible in the shadows, no color grading. Unposed, nobody looking at the camera.
The controls: imperfection. “Slightly imperfect framing,” “one element clipped,” and “nobody looking at the camera” are what break the centered, symmetrical, obviously-composed look that generated images default to.
Part 3: Cinematic
Photorealistic cinematic shot
Prompt to try
Anamorphic wide shot, 2.39:1. A lone figure at the far end of a fluorescent-lit underground car park, seen from behind. Practical light sources only — ceiling strips and one flickering fixture. Deep shadows with detail retained, teal-leaning shadows and warm sodium highlights. Slight lens flare from the flickering unit. Subject occupies a small portion of the frame. Film grain, no digital sharpening.
The controls: aspect ratio, “practical light sources only,” and scale relationship between subject and frame. Cinematic is mostly a lighting and framing decision, not a filter.
Realistic film still
Prompt to try
A single frame from a 1970s European drama. Two people at a kitchen table, mid-conversation, neither looking at the other. Window light from the right, heavy shadow on the left half of the frame. Muted, slightly faded color stock with warm cast and lifted blacks. Static camera, eye level, natural composition. Everyday clutter on the table. No text.
The controls: “mid-conversation, neither looking at the other” produces the sense that you have interrupted a scene. Narrative beats visual description for this style.
Part 4: Interiors and renovation
Before-and-after renovation
Prompt to try
Using the uploaded photo of this room, keep the exact architecture — window positions, ceiling height, doorway, and camera angle unchanged. Reimagine it as a warm minimalist living room: pale oak flooring, lime-plaster walls in soft white, a low linen sofa, one large plant, no clutter. Same time of day and light direction as the original.
The controls: “keep the exact architecture” and “same camera angle” are the difference between a before-and-after pair and two unrelated pictures. Uploading the actual photo matters more than any adjective.
Interior design exploration
Prompt to try
Same room, same angle, three variations: one Japandi with warm neutrals and low furniture, one mid-century with walnut and burnt orange, one contemporary industrial with blackened steel and concrete. Keep the layout and light identical across all three so I can compare them fairly.
Quick reference
| Style | The word that controls it | Fix when it misses |
|---|---|---|
| Cartoon poster | Era and print process | Change the decade |
| Anime | “Background art” vs “character” | Add “cel shading, visible brushwork” |
| Fashion editorial | Lighting setup | Add “no beauty retouching” |
| Street photo | Imperfection cues | Add “clipped by the frame edge” |
| Cinematic | Aspect ratio and practical light | Reduce subject scale in frame |
| Film still | A narrative beat | Describe what just happened |
| Renovation | “Keep the architecture” | Upload the real photo |
| Interior variants | “Identical layout and light” | Generate all variants in one request |
Where this fits in ChatUp
Image Gen covers text-to-image and can use a reference image when the selected image model supports it. Animate Photo takes a still and produces a short clip. For interior work, treat generated output as a concept rather than an accurate plan of the room.
For the campaign-focused version of this, see how to create marketing images with AI.
Frequently asked questions
Why do my images look obviously AI-generated?
Usually three things: perfect symmetry, over-smooth skin or surfaces, and lighting with no identifiable source. Adding imperfection, texture, and a named light setup fixes most of it.
How do I get the same style across several images?
Save the style portion of your prompt as a reusable block and paste it into every generation. Vary only the subject. Generating variations from an approved image beats writing a new prompt each time.
Can I generate an image in a specific artist’s style?
Many tools restrict named living artists, and the ethics are contested regardless. Describing the technique — medium, palette, brushwork, era — gets you most of the way and avoids the problem entirely.
Why does the same prompt give different results each time?
Generation is stochastic by design. If you need consistency, iterate on one image rather than regenerating, and request variations in a single request so they share a starting point.
Style is a list of decisions
Every one of these recipes is the same idea: replace the adjective with the decisions the adjective implies. “Cinematic” becomes an aspect ratio, a lighting rule, and a scale relationship. Once you can see the decisions inside a look, you can produce it — and, more usefully, fix it when it misses.
Try it in ChatUp
Turn this guide into a workflow.
Run the prompts above against the model that suits the task, keep the useful context across chats, and pick it back up on any device.
Try for Free