AI image generators can produce remarkable results, but the quality of the output depends heavily on how clearly you describe what you want.
A prompt such as:
Make an anime picture of a woman in a greenhouse.
communicates the basic idea, but it leaves most of the creative decisions to the AI.
What kind of anime aesthetic should it use? What is the woman doing? How is the scene composed? What time of day is it? What colours dominate the image? Which details must remain accurate?
A stronger prompt answers those questions deliberately.
To make this easier, I created the SOXX Prompt Atlas—a responsive reference guide containing a practical prompt-building framework and 67 reusable visual styles.
Explore the SOXX Prompt Atlas
Open the responsive prompt guide
The guide lets you:
Search and filter different visual styles
Copy reusable prompts
See finished prompt examples
Learn what each part of a prompt controls
Add further aesthetic direction
Adapt prompts for different AI image generators
The most useful styles appear at the beginning of the library, followed by a broader collection covering illustration, photography, cinema, comics, printmaking, 3D design and specialist visual treatments.
The Anatomy of a Strong Image Prompt
A reliable image prompt can be structured like this:
[SUBJECT OR SCENE] + [ENVIRONMENT] + [STYLE OR MEDIUM] + [STYLE TRAITS] + [COMPOSITION OR CAMERA] + [LIGHTING] + [COLOUR PALETTE] + [MOOD] + [MATERIALS OR TEXTURE] + [FINISH] + [CONSTRAINTS]
You do not need to use every component in every prompt. The framework is there to help you identify what information is missing.
Let’s examine each part.
1. Subject or scene
Describe the main person, creature, object or event. Include the action when something is happening.
Weak:
A botanist.
Stronger:
A young botanist examining a luminous orchid.
The second version gives the image generator a subject, an action and an important visual object.
2. Environment
Explain where and when the scene takes place.
You can include:
Location
Time of day
Weather
Historical period
Background activity
Supporting objects
Example:
Inside a rain-soaked rooftop greenhouse at night.
3. Style or medium
Describe the broad visual treatment and how the image should appear to have been created.
Examples include:
Cinematic anime illustration
Traditional watercolour painting
High-end editorial photography
Modern comic-book artwork
Polished stylised 3D animation
Charcoal drawing
Risograph print
Pixel art
Oil painting
Scientific illustration
A named style alone may not be enough. Different generators can interpret the same label differently, so it helps to describe the visible characteristics of that style.
4. Style traits
These are the qualities that make the chosen visual style recognisable.
For anime, you might specify:
Clean linework, expressive eyes, detailed hair and cel shading.
For a family-friendly animated-film aesthetic:
Appealing rounded forms, expressive faces, tactile materials, readable silhouettes and warm emotional storytelling.
For a comic-book treatment:
Bold ink contours, controlled cross-hatching, graphic shadows and saturated accent colours.
This approach is especially useful when you want an aesthetic associated with a particular studio, comic tradition or creative era. Describe the observable qualities instead of depending entirely on a company or franchise name.
5. Composition and camera
Composition determines how the viewer experiences the subject.
Useful instructions include:
Close-up
Medium shot
Wide establishing shot
Eye-level view
Low-angle view
Overhead view
Centred composition
Asymmetrical composition
Shallow depth of field
Layered foreground
Strong negative space
Vertical poster composition
Widescreen cinematic framing
Example:
Low-angle medium shot with layered foreground leaves and shallow depth of field.
6. Lighting
Lighting can completely change the emotion and perceived quality of an image.
You can describe:
Light source
Direction
Hardness or softness
Colour
Contrast
Time of day
Atmospheric effects
Example:
Teal and magenta rim lighting with a warm key light on the face.
Other useful lighting descriptions include:
Soft directional window light
Golden-hour backlighting
Dramatic chiaroscuro
Cold industrial practical lighting
Warm diffused daylight
Volumetric god rays
Hard side lighting
Large studio softbox
Radiant stained-glass backlighting
7. Colour palette
Naming a limited palette helps produce a more visually coherent image.
Example:
Teal, magenta, emerald and warm amber.
You can also describe the palette by character:
Muted earth tones
Saturated tropical colours
Sun-faded vintage colours
Monochrome with one red accent
Black, gold and deep teal
Soft pastel colours
High-contrast red, yellow, green and black
8. Mood
State the emotional feeling the image should communicate.
Examples include:
Hopeful
Mysterious
Joyful
Intimate
Tense
Nostalgic
Playful
Ominous
Meditative
Energetic
Luxurious
Unsettling
Example:
Mysterious, hopeful and wondrous.
9. Materials and texture
Surface details make an image feel more tangible.
Example:
Wet glass, waxy leaves, glowing pollen and tactile fabric.
Other useful texture descriptions include:
Visible paper grain
Layered brushstrokes
Fingerprint-textured clay
Cracked masonry
Woven fibres
Polished metal
Frosted glass
Rough charcoal
Coarse halftone dots
Embroidered thread
Handmade paper edges
10. Detail level and finish
Tell the generator how polished or simplified the final image should appear.
Examples include:
High-detail key visual
Feature-film-quality render
Clean scalable vector finish
Handmade illustrated finish
Rough concept sketch
Publication-ready scientific plate
Premium advertising retouch
Weathered screen-print texture
11. Non-negotiable details
These are the elements the image generator must preserve.
Example:
Preserve the orchid’s six petals and the botanist’s round glasses.
This section becomes especially important when generating:
Recurring characters
Real people
Products
Uniforms
Logos
Packaging
Architecture
Cultural artefacts
Scientific subjects
If you are using a reference image, explain whether it controls identity, clothing, pose, composition, product accuracy or visual style.
12. Unwanted elements
Instead of writing a long list of vague negatives, identify a few specific failures that would damage the image.
Example:
Avoid extra fingers, duplicated flowers, illegible labels and background clutter.
Negative instructions work best as cleanup. They should not replace positive creative direction.
A Complete Prompt Example
Here is the full prompt assembled from the components above:
A young botanist examining a luminous orchid inside a rain-soaked rooftop greenhouse at night, rendered as a cinematic anime illustration with clean linework, expressive eyes, detailed hair and cel shading. Low-angle medium shot with layered foreground leaves and shallow depth of field. Teal and magenta rim lighting with a warm key light on the face; teal, magenta, emerald and warm amber palette; mysterious, hopeful and wondrous mood. Wet glass, waxy leaves, glowing pollen and tactile fabric; high-detail polished key-visual finish. Preserve the orchid’s six petals and the botanist’s round glasses. Avoid extra fingers, duplicated flowers, illegible labels and background clutter. Use a vertical 4:5 composition and the supplied character reference for identity.
The final prompt reads naturally, but every sentence performs a specific job.
Do Different AI Image Generators Need Different Prompt Structures?
Yes—but you do not need to rewrite your entire creative idea for every platform.
The safest approach is to maintain a portable natural-language prompt and then add controls supported by the particular generator you are using.
Conversational image generators
These generally respond well to complete sentences and compact creative briefs.
Start with a natural-language description. If the first result is close, revise one visual variable at a time.
For example:
Keep the character and composition unchanged, but replace the neon lighting with soft sunrise light.
Tag-oriented diffusion interfaces
Some diffusion systems respond strongly to important keywords and model-specific vocabulary.
Place the primary subject and most important aesthetic instructions early. If the interface provides a separate negative-prompt field, use it instead of mixing a large negative list into the main description.
Model weights, samplers, guidance settings and specialised vocabulary can vary. Do not assume settings copied from one model will behave the same way in another.
Parameter-based generators
Some systems allow parameters for:
Aspect ratio
Seed
Stylisation
Variation
Reference strength
Image weighting
Write the understandable prompt first. Add the supported parameter syntax afterwards.
Keep a clean copy of the original prompt because parameter formats can change.
Design-suite generators
Design applications may provide separate controls for:
Medium
Composition
Colour
Aspect ratio
Reference images
Commercially safe presets
Use those interface controls where appropriate rather than repeating everything inside the written prompt.
Reference-image workflows
Tell the generator exactly what each reference controls.
For example:
Use Reference 1 for facial identity, Reference 2 for the clothing design and Reference 3 only for the colour palette and print texture.
Without this direction, the generator may blend the references unpredictably.
Studio, Comic and Recognisable Aesthetics
Sometimes we know the look we want because it reminds us of a famous animation studio, comic publisher, film period or artist.
However, simply entering a studio name does not always produce a consistent result. Different models can interpret those names differently, and some platforms may restrict them.
A more reliable method is to translate the reference into visible qualities.
Instead of relying only on:
Disney Pixar style
try something like:
Polished family-friendly stylised 3D animation, appealing rounded character proportions, large expressive eyes, tactile hair and fabric, warm emotional storytelling, soft cinematic lighting, readable silhouette and feature-film-quality rendering.
Instead of relying only on:
Superhero comic style
try:
Modern superhero comic-book illustration with bold ink contours, controlled cross-hatching, dramatic anatomy, graphic shadows, saturated accent colours, heroic rim lighting and a foreshortened cover composition.
This makes the prompt more portable and gives you greater creative control.
Styles Included in the SOXX Prompt Atlas
The guide contains 67 visual directions, including:
Cinematic anime
Family-friendly stylised 3D animation
Photorealistic editorial photography
Cinematic film stills
Watercolour
Modern comic-book illustration
Children’s picture-book illustration
Studio portrait photography
Fantasy and science-fiction concept art
Minimal flat vector illustration
Oil painting
Graphic-novel noir
Product advertising
Isometric 3D
Manga
Retro anime
Chibi illustration
Gouache
Coloured pencil
Charcoal
Ink wash
Pastel
Art Nouveau
Art Deco
Pop art
Surrealism
Psychedelic posters
Low-poly illustration
Newspaper comics
European adventure comics
Underground comix
Risograph
Vintage travel posters
1950s advertising
1980s airbrush art
Y2K futurism
Vaporwave
Synthwave
Pixel art
Documentary photography
Street photography
Fashion editorials
Food photography
Architectural photography
Macro photography
Aerial photography
Black-and-white fine art
Instant-film snapshots
Analog 35mm photography
Horror film stills
Claymation
Paper cutouts
Felt craft
Interlocking toy-brick worlds
Voxel art
Glassmorphism
Papercraft
Embroidery
Stained glass
Ceramic relief
Technical blueprints
Scientific illustration
Infographic posters
Minimalist luxury
Dark-fantasy engraving
Botanical plates
Each entry includes a reusable prompt, a description, additional aesthetic guidance and a finished example.
A Simple Workflow for Better Results
When creating your next AI-generated image:
Define the subject and action.
Establish the environment.
Choose one primary visual style.
Describe the visible traits of that style.
Direct the composition and camera.
Set the lighting, palette and mood.
Add important materials and textures.
State which details must remain accurate.
Identify a few specific failures to avoid.
Add platform-specific controls only when supported.
Generate several variations.
Select the strongest composition and refine one issue at a time.
Avoid loading the first prompt with every adjective you can think of. Clarity is more valuable than length.
Final Thought
A good AI image prompt is not a magic spell.
It is art direction.
The goal is not to discover a secret combination of keywords. The goal is to communicate your visual intention clearly enough that the image generator can make useful decisions.
Describe the subject. Establish the world. Direct the camera. Shape the light. Define the aesthetic. Protect the details that matter.
Then generate, evaluate and refine.
SOXX — Mellah Kilangit
Digital Kanaka · Port Moresby, Papua New Guinea
Good for nothing Loafer of the Internet.