IMAGEKA's style presets are not filters. Each one appends a fixed string of words to the end of your prompt, and the preset also helps IMAGEKA decide which of its image models draws the picture. Once you know exactly what those words are and where each preset routes, you can predict what a preset will do, choose the right one for your subject, and recognise when it is better to switch to Default and describe the style yourself.
- A preset adds extra prompt text after yours and steers which model is used: Photorealistic or Vintage with a person, product or food goes to the photorealism model; Anime to the anime model; Oil Painting, Watercolor, Sketch and Pixel Art to the painterly model; the rest to the fast general model.
- Photorealistic is selected by default and adds an 85mm portrait lens with shallow depth of field, which suits people and close-ups but not wide landscapes or interiors.
- If your prompt already names a medium, lens or palette, choose Default so the preset does not contradict it.
- Presets work best with subjects that match the medium they describe: Pixel Art for game objects, Watercolor for soft natural subjects, Minimalist for simple graphic ideas.
How presets actually work
When you type "a lighthouse on a cliff" and choose Watercolor, the model receives "a lighthouse on a cliff, watercolor illustration, soft washes, paper texture, wet on wet technique". The preset choice is also one of the signals IMAGEKA uses to pick a model, so Watercolor sends this prompt to the painterly model. You could type those words yourself with Default selected and get essentially the same result, because the word "watercolor" in a prompt routes it the same way.
This matters for two reasons. First, the preset text is written for the common case, so it will sometimes contain a phrase that does not suit your image. Second, the model reads the whole prompt together, so preset words and your own words interact. If you write "a black and white photo" and the preset adds "vibrant colors", the model has to reconcile two instructions, and the result is usually a compromise that satisfies neither. The how it works guide explains why the text encoder weighs every phrase rather than following the first or last one.
Presets push the style firmly, but they rarely override a strongly described subject.
Which model each preset routes to
IMAGEKA runs several open-source image models and picks one per request. You cannot choose the model directly, but the preset is the strongest lever you have. This table shows the kind of model each preset usually leads to:
| Preset | Model it routes to | Notes |
|---|---|---|
| Photorealistic, Vintage | Photorealism model, when the prompt features a person, portrait, product or food | Other subjects (landscapes, cities, interiors) stay on the fast general model |
| Anime | Anime model | Also chosen with Default if you write "anime", "manga" or "chibi". Responds well to comma-separated tags |
| Oil Painting, Watercolor, Sketch, Pixel Art | Painterly model | Also chosen for words such as watercolour, charcoal, gouache, impasto, linocut or pixel art |
| Digital Art, Concept Art, 3D Render, Minimalist | Fast general model | Unless your own words point to a specialist (for example "anime" or "watercolour") |
| Default | Decided by your words alone | Photo words plus a person, product or food still reach the photorealism model |
One rule overrides the preset: text in quotes, or words such as poster, logo, label, icon, sticker or typography, send the prompt to the fast general model whatever preset is selected, because it is the most reliable at lettering and layout. The three specialists use the Negative Prompt field and get sensible default negatives (bad hands, blurry, watermark and similar) added automatically; the general model largely ignores negatives. The specialists also produce at least 768×768, so a 512×512 request is generated at 768×768 by them.
The twelve presets, one by one
Each section below quotes the exact text the preset appends, explains what each phrase contributes, and gives a prompt you can open with that preset selected.
Default
Adds: nothing.
Default sends your prompt exactly as written. Choose it whenever you have described the style yourself (your own style words still steer the model choice), whether that is "linocut print", "shot on a 24mm lens, everything in focus" or "flat vector illustration in navy and coral". It is also the right choice for text-heavy designs like posters, where style words compete with the lettering for the model's attention; quoted text sends the prompt to the general model, which handles lettering best. Many experienced users leave it on Default permanently and keep a few favourite style phrases to paste in.
Photorealistic
Adds: "photo taken on Sony A7R, 85mm lens, shallow depth of field, photorealistic, ultra sharp"
The camera name signals a high-resolution full-frame photograph rather than a phone snap. "85mm lens" is the important phrase: it is a classic portrait focal length, which means a tight, flattering frame with little background visible. "Shallow depth of field" blurs the background, and "ultra sharp" keeps the subject crisp. Together they produce a polished portrait or product look. When the prompt features a person, product or food, this preset also routes it to IMAGEKA's photorealism model, which is strongest on skin texture and natural light.
Best for: head-and-shoulders portraits, products, food close-ups, macro nature, single objects.
Pitfalls: because it is selected by default, it is easy to forget it is there. On a wide landscape, a room interior or a street scene it fights your intent: you ask for a sweeping view and get a cropped frame with a blurred background. For those, choose Default and write "wide-angle lens, everything in focus" yourself. The lighting and camera guide covers which lens terms produce which framing.
Digital Art
Adds: "digital illustration, vibrant colors, sharp clean lines, polished painterly finish"
"Digital illustration" moves the image away from photography. "Vibrant colors" raises saturation, "sharp clean lines" gives defined edges, and "polished painterly finish" adds smooth rendered shading. The result is the glossy look common on game cards, book covers and fan-art style illustrations. This preset is handled by the fast general model.
Best for: characters, creatures, book-cover concepts, colourful scenes, thumbnails that need to stand out.
Pitfalls: "vibrant colors" clashes with muted or monochrome palettes. If you want a dusky, restrained illustration, use Default and describe the palette. It can also look generic when the subject is generic; give it a specific character and setting.
Oil Painting
Adds: "oil on canvas, thick impasto brushwork, museum quality, old masters style"
"Oil on canvas" sets the medium. "Thick impasto brushwork" asks for visible raised paint strokes. "Museum quality" pushes towards finished, serious work, and "old masters style" pulls the palette and lighting towards classical European painting: warm earth tones, dark backgrounds and directional light. Oil Painting routes to the painterly model.
Best for: portraits with a classical feel, still lifes, pastoral landscapes, historical scenes, pet portraits as gifts.
Pitfalls: "old masters style" darkens and warms everything, so modern or brightly coloured subjects such as a neon city can look muddy. Impasto and very fine detail also pull in opposite directions; expect loose brushwork rather than crisp small features.
Watercolor
Adds: "watercolor illustration, soft washes, paper texture, wet on wet technique"
"Soft washes" gives broad translucent areas of colour, "paper texture" adds visible grain, and "wet on wet technique" describes paint applied to damp paper, which creates soft bleeding edges. The overall effect is light, airy and loose.
Best for: botanical studies, food and menu illustrations, gentle landscapes, travel sketches, greeting cards, children's book scenes.
Pitfalls: wet-on-wet blurs edges, so it is a poor choice when you need sharp detail, small text or precise architecture. Dark, high-contrast scenes also fight the medium, which is naturally pale.
Anime
Adds: "anime style, hand-drawn cel shading, clean linework, soft painted background, vibrant"
"Cel shading" means flat areas of colour with hard-edged shadows, as in traditional animation. "Clean linework" gives crisp outlines on characters. "Soft painted background" separates sharp characters from a more atmospheric painted setting, which is a defining trait of the style. "Vibrant" lifts saturation.
The Anime preset sends your prompt to IMAGEKA's anime model. Natural sentences work, but this model also responds well to short comma-separated tags, so "1girl, long black hair, red scarf, winter coat, snowy street, night" is a perfectly good prompt here. A few character tags followed by a sentence describing the scene often gives the best of both.
Best for: characters in everyday or dramatic scenes, scenic skies, city streets, school and slice-of-life settings, avatar concepts.
Pitfalls: words in quotes or design words such as "poster" or "logo" send the prompt to the general model instead, so keep them out of anime prompts. Do not add the name of a specific studio or series; describe the scene and let the preset supply the look. Very photographic words, like lens names or "skin pores", work against it. Hands in complex poses remain error-prone, as they are in every style.
Concept Art
Adds: "cinematic concept art, epic composition, dramatic lighting, highly detailed environment"
This is the preset most focused on the setting. "Cinematic" suggests a wide, film-like frame. "Epic composition" encourages large scale and depth, "dramatic lighting" adds strong contrast and light shafts, and "highly detailed environment" tells the model to spend its detail on the world rather than on a single figure.
Best for: fantasy and sci-fi landscapes, cities, ruins, spaceships, game and story worldbuilding, dramatic book-cover backgrounds.
Pitfalls: "epic" and "dramatic" will inflate a modest subject. A cosy kitchen becomes a cavernous hall with shafts of light. It is also a poor fit for close-up portraits, where the environment emphasis pulls the camera back. Use 1024×768 so the composition has room.
Sketch
Adds: "detailed pencil sketch, graphite on paper, expressive hand-drawn lines"
"Graphite on paper" fixes the medium as monochrome grey pencil. "Detailed" asks for a finished drawing rather than a few rough marks, and "expressive hand-drawn lines" allows visible, varied strokes and hatching. Like the other traditional-media presets, Sketch routes to the painterly model.
Best for: architecture, portraits with character, animals, objects, storyboards, a hand-drawn look for posters and articles.
Pitfalls: colour words in your prompt either get ignored or produce tinted pencil effects; if you want colour, write "coloured pencil drawing" with Default instead. Very busy scenes can turn into a tangle of lines; one clear subject works best.
3D Render
Adds: "photorealistic 3D render, octane render, ray tracing, subsurface scattering, 8K"
"3D render" and "octane render" (a well-known rendering engine) push towards the clean, computer-generated look of product visualisation. "Ray tracing" encourages accurate reflections and shadows. "Subsurface scattering" describes light passing slightly into materials like wax, skin, jade or soap, which gives them a soft glow. "8K" contributes little but does no harm. 3D Render is handled by the fast general model.
Best for: product mock-ups, cute stylised characters and toys, 3D lettering, abstract shapes, icons with depth, glossy or translucent materials.
Pitfalls: despite the word "photorealistic", results look rendered rather than photographed: surfaces are often too clean and perfect. For a convincing real-world photo, use Photorealistic or Default with photographic language. It also suits stylised characters better than realistic people, who can look waxy.
Minimalist
Adds: "minimalist design, clean negative space, simple geometric forms, flat color"
"Clean negative space" leaves large empty areas, "simple geometric forms" reduces objects to basic shapes, and "flat color" removes gradients and texture. This preset changes the image more than most, turning almost any subject into a graphic poster-like design.
Best for: posters, wallpapers, icons, book and podcast covers, slide backgrounds, headers that need room for text.
Pitfalls: it strips out detail, so prompts that list many objects or describe textures get simplified or ignored. Keep your prompt short: one subject, one or two colours, one placement. For more on using empty space well, see colour and composition.
Vintage
Adds: "vintage film photograph, Kodak Portra 400, light leak, nostalgic warm tones"
"Vintage film photograph" moves away from crisp digital capture towards grain and softer contrast. Portra 400 is a colour negative film known for warm, gentle skin tones, so naming it shifts colour in that direction. "Light leak" adds orange or red flares at the frame edges, and "nostalgic warm tones" warms the whole image.
Best for: candid people, summer and travel scenes, cars and roadside places, family-album or retro brand moods.
Pitfalls: light leaks appear somewhat randomly and can cover an important part of the frame; generate several. Clean product shots and cold, blue scenes look wrong with the warm cast. Modern objects such as phones can look odd in a "vintage" frame, which may or may not be what you want.
Pixel Art
Adds: "16-bit pixel art, retro game sprite, limited color palette, crisp pixels"
"16-bit" points at the style of home consoles from the early 1990s, with more colour than 8-bit but still blocky. "Retro game sprite" frames the subject as a game object or character, "limited color palette" reduces the number of colours, and "crisp pixels" asks for hard square edges without blur.
Best for: game items, small characters, cosy scenes, retro-themed social posts, avatars.
Pitfalls: the model imitates pixel art; it does not produce a true low-resolution grid. Pixel sizes may be inconsistent and edges slightly soft, so for actual game assets, downscale and clean up in a pixel editor. "Sprite" tends to isolate a single subject, so large detailed scenes can come out confused.
Comparison table
| Preset | Best for | Avoid for |
|---|---|---|
| Default | Prompts that describe their own style; posters with text | Very short prompts where you want a finished look quickly |
| Photorealistic | Portraits, products, food and macro close-ups | Wide landscapes, interiors, street scenes |
| Digital Art | Characters, creatures, colourful covers | Muted, monochrome or documentary looks |
| Oil Painting | Classical portraits, still lifes, pastoral scenes | Neon, modern or brightly coloured subjects; fine detail |
| Watercolor | Botanicals, food illustrations, gentle landscapes | Sharp detail, text, dark high-contrast scenes |
| Anime | Characters in scenes, skies, everyday moments | Photographic detail; named studios or series |
| Concept Art | Large fantasy and sci-fi environments | Close-up portraits, small cosy scenes |
| Sketch | Architecture, animals, objects, storyboards | Anything that depends on colour |
| 3D Render | Product mock-ups, stylised toys, 3D type | Believable real-world photos and realistic people |
| Minimalist | Posters, wallpapers, covers, icons | Detailed, multi-object or textured scenes |
| Vintage | Candid people, travel, retro moods | Clean product shots, cold blue scenes |
| Pixel Art | Game items, small characters, cosy scenes | Large busy scenes; production-ready sprites |
When to choose Default and write the style yourself
Presets are a shortcut for a dozen common looks. Choose Default when:
- You want a medium not on the list. Linocut, risograph, gouache, charcoal, stained glass, paper-cut, embroidery and claymation all work well when written into the prompt.
- Part of a preset is wrong for you. If you want a photograph but with a wide lens and everything in focus, Photorealistic's "85mm, shallow depth of field" is the opposite of what you need. Write the photographic terms you do want.
- Your image contains text. Extra style words dilute attention on the quoted words. Posters and signs generally come out cleaner on Default with a short style description.
- You need a consistent series. Writing the same style sentence into every prompt makes the look explicit and easy to tweak, which helps keep a social feed or a set of blog headers consistent.
Combining and overriding presets
You can select only one preset at a time, but because a preset is just text, you can combine styles by hand. Choose Default and paste the text of two presets, then edit out the phrases that conflict. For example, Watercolor plus Sketch becomes "pencil sketch with soft watercolor washes, graphite lines on textured paper", a classic ink-and-wash look that neither preset gives alone.
You can also keep a preset and steer it. Your own words come first in the final prompt and the model takes the whole sentence into account, so adding "muted earthy palette" to a Digital Art prompt will soften but not completely remove "vibrant colors". If a phrase in the preset keeps winning, that is the signal to switch to Default and write the style without it.
Do not rely on the Negative Prompt field to cancel preset words. The photo, anime and painterly models do respond to negatives, but a negative fighting a phrase in the positive prompt usually ends in a compromise, and the fast general model largely ignores negatives anyway. Removing the conflicting phrase from the prompt itself is the reliable way.
Why presets clash with style words in your prompt
The most common cause of a muddled image is contradictory style information. A few clashes appear again and again:
a black and white documentary photo of a market
With Digital Art selected, the prompt ends with "digital illustration, vibrant colors". The result tends to be a half-painted image with patches of colour.
a black and white documentary photo of a busy market, 35mm lens, natural light, film grain
With Default selected, there is one consistent description and the model follows it.
Others to watch for: "watercolour painting" with the Photorealistic preset still selected (you get a photo of a painting, or a strange hybrid); "wide aerial view" with Photorealistic's 85mm lens; "minimal flat icon" with Concept Art's "highly detailed environment". In each case, the fix is the same: pick the preset whose words agree with yours, or pick Default.
Photorealistic is pre-selected when you open the generator. If you paste a prompt that already names a style, check the style chip before you generate. The prompt library sets the right preset automatically when you use its Try it button.
Frequently asked questions
Does choosing a preset use a different AI model?
Often, yes. IMAGEKA uses several open-source models and picks one automatically based on your prompt and preset. Photorealistic or Vintage with a person, product or food goes to the photorealism model, Anime to the anime model, and Oil Painting, Watercolor, Sketch and Pixel Art to the painterly model. Everything else, and anything with quoted text, goes to the fast general model. See which model each preset routes to.
Can I choose the model myself?
Not directly; the choice is automatic. The preset and words such as "photograph", "anime" or "watercolour" are how you steer it.
Why is Photorealistic selected by default?
Many people start with photographic subjects such as people, products and food, and the preset gives those a polished finish without any extra effort. It is not ideal for every subject, so change it to Default or another preset when it does not fit.
Can I see the full prompt that is sent, including the preset text?
The preset text is listed on this page for every style. Your prompt plus that text, separated by a comma, is what the model receives.
Why do I get different results with the same prompt and preset?
The seed is random on every run, and there is no seed control in the interface. The style stays recognisably the same, but composition and details change. Generate two or four images and choose the best.
Can I ask for the style of a specific artist?
We recommend describing the qualities you like instead: the medium, palette, brushwork, era and mood. That gives you more control, and it avoids imitating a living artist's personal style, which raises ethical and sometimes legal concerns.
Does the "Enhance prompt" button change the style?
Slightly. It adds a few phrases based on keywords in your prompt, such as "detailed eyes" for portraits, plus "dramatic lighting" if you have not mentioned light. If a preset is selected, it also writes that preset's first phrase (for example "watercolor illustration") into your prompt, so the style is then stated twice when you generate. Read what it added and delete anything that conflicts with the look you want.