The fabulous guide to styles

[Fuka]

Hey you’re here most likely to learn about how styles work! Great! Be prepared to read a wall of text after a wall of text. Some images included™️
This tutorial assumes you have actually read the advanced tutorial

Let us begin!

By reading previous relevant tutorials you probably have a general idea how to structure your prompts. The thing is physical, location/background and clothing tags are more of a generalized suggestion than an actual necessity.

First of all lets see how different sets of prompts affect image generation
Here is my private character Yoru
Very Long Side Bangs, Pink Eyes, Blue Lips, Pink Short Hair, Full Lips, Big Hips, Curvy Figure, Slender Waist, Petite Build, Fair Skin, Thick Thighs, Soft Curves, Adult Woman

Yoru but with stylized Bakground
Background: Cinematic lighting;Film look, Low key light;Dark background, Directional light/Golden hour, Natural color;Color graded, Tonal contrast;Moody blacks, Depth of field/Bokeh effects, Lens bloom/Anamorphic flare, Real lens distortion;HDR, Sharp focus;Clarity, Skin shading/Ambient occlusion, Filmic soft shadows/Film grain

Here are physical prompts for Aria
Buxom Body Type :2, Round Feminine Face :2, Anime Style Eyes :2, Soft Details, Detailed_hair :2, Tall, Soft Feminine Face, Thicc Thighs, Detailed_eyebrows:2, Detailed_eyes:2, Detailed Clothing :2, Adult_woman :2, Intricate_hair_details :10, Blonde Hair, Long Eyelashes :2, Detailed Clothing Texture :2, Blue Very Detailed Eyes :2, Classic Anime Style :2

Prompt strength 11 on A-OG

Here is her with an example of style prompts
Clothing: Cinematic lighting;Film look, Low key light;Dark background, Directional light/Golden hour, Natural color;Color graded, Tonal contrast;Moody blacks, Depth of field/Bokeh effects, Lens bloom/Anamorphic flare, Real lens distortion;HDR, Sharp focus;Clarity, Skin shading/Ambient occlusion, Filmic soft shadows/Film grain, Photographic framing
Background: Cowboy_shot:2, Film_grain:1.1, Cinematic_composition:1.4, Depth_of_field:1.5, Dramatic_angle:1.2, Close-Up:2, Detailed_eyes:1.4, Soft_lighting:1.2, Highres:1.2, Fine_details:1.3, Atmospheric:1.2, Portrait:1.5, Looking_at_viewer:1.4, Bokeh:1.3, Sharp_focus:1.4, Upper_body:1.3, Blurred_background:1.5, Masterpiece:1.2

Quite overwhelming isn’t it!

That is what we are solving in this tutorial.

What we are doing with those tags is giving the AI more information about how we want pictures to look like.

This can be done with great many things be it physical tags, background, clothing or loras.
Prompting in itself isn’t really complicated if you know where to look for resources. I strongly suggest repositories with tags like danbooru
Image composition is everything you really need to create pictures like this one

ADDBASE (epic, grand, scene)+++, (solo, alone)+++, (panorama)++, (chinese theme, ancient china)+++, (lonely, alone)+++, Muted_colors+++

Main things you need to understand when writing your own prompt:

• Be extremely precise,
• Be mindful of redundant tags (tags AI does not understand)
• Always pay attention to weights (if you put too much or in pointless places the math will break and it will start showing up artifacts)
• Have an outcome in mind.

With this behind us let us go to another topic:

Let’s talk about style

Now to explain what makes the overall artstyle and what affects it.
The biggest thing for you as a creator is understanding how to properly describe an image. This can be done by stating what parts of image composition you have in mind.
Below you will find a table showing just that in action.

Term	Explanation	Example	Example Image
Style / Stylization	Overall look	(Realistic_style)2
Shading / lighting	Defining the volume and depth	(realistic_shading)2
line-art	Definition of edges and forms	(clean_line-art)2
Proportions	How subjects are depicted	(anime_proportions)2
Texture / detail	The level and type of surface details	(highly_detailed)2
composition / framing	How things are arranged	(isometric_compostion)2
Simulation	what medium does the picture simulate	(Oil_painting simulation)2

Using this knowledge we can actually start constructing a basic style.
example:

(realistic_style)2 (painterly_shading)2	(realistic_shading)2 (painterly_style)2

Now after seeing it you might be asking why didn’t I use a , between the keywords and here is an example of what happens when you add them ⁠

(realistic_shading)2 ,(painterly_style)2	(realistic_shading)2 (painterly_style)2

The picture changed A LOT and that’s because of how the ai sees your prompt

Case 1 - using a ,	The user wants a realistic shading > now make it painterly style

Case2 - not using a ,	The user wants a painterly styled picture with realistic shading

Weights

There are a lot of misconceptions about how to use weights on the site. Soulkyn does not let you use negative prompts this means that as long as you use a tag it will appear. The base value is 1. If you lower it to 0.001 it is still higher than 0. If you don’t want stuff to appear do not add it.

This means that if you add Photorealistic anime style it’s default weight is 1. You can modify it to be like this in physicals Photorealistic anime style :0.4 or Photorealistic style :1.4 it just changed how much the tag influences the outcome. Due to normalization of tags even if they have :100 included they will be brought down to around 2.

Here is an example of one of my many Fukas on Anime-Hyper with variations of photorealistic.

No	0.4	1	2

As you can clearly see the weights in physicals actively change how the style changes towards more realistic.
This honestly applies to anything you are doing with your kyn, be it sizes of body parts or etc.

---

Creating the Style

Now with the afromentioned knowledge we can start creating our first style prompts that will be saved as background presets.

Few styles we will be making right now:

• Old Anime Style
• Photorealistic
• Cinematic
• Epic

For this I will be using StarFuka with Anime-Hyper

We won’t be creating anything very intricate because this isn’t our goal. Our goal is to learn and it is best done with small steps.

Old Anime Style

What is the first thing that comes to mind when we say “old anime”. For me it is bold black lineart. The 1980s perfect black outline of the characters, the simple style, Flat colors etc. By now you might already be grasping what I mean when I mention those. They are a form of tags in themselves.

1980s_anime_style, Bold Lineart, Thick lineart, Flat colors, Simple_anime_style, classical anime style, old anime style, vector shading

Now let us analyze what tags do what:

1980s_anime_style, Simple_anime_style, classical anime style, old anime style - Those tags are here to better tell AI what sort of anime style we want.

Bold / Thick Lineart - This describes what sort of lineart we want

Flat Colors - Flat colors are colors that don’t really blend, don’t have depth due to shading, light etc. They are flat.

Vector Shading - Vector shading is simple shading that is just a patch of darker color not blended with the rest of the image.

Photorealistic

What comes to mind when you hear Photorealism?
It is mostly realistic details. Realistic shading, realistic style, realistic hair, eyes and all of that. Let us describe it for the AI.

Realistic Hair, Realistic Eyes, Photorealistic Style :2, Very Detailed, Best, Realistic, Photorealistic Shading, Realistic Character Style, Photorealistic Anime Style, Realistic Shading :2, Realistic Shadows

I think the tags for this one are self explanatory but:

Realistic/photorealistic XYZ - We want to tell AI which parts of the picture we want to be realistic. The more detailed description the better the result to a certain point.

Best/Detailed - We use this to reinforce the quality.

Cinematic

On the other hand this one is tricky. What do you imagine when you hear phrase “Cinematic shot”? For me it is lens flare, blurry background, sharp focus, bokeh, flying particles, basically all nine yards when it comes to “film and photography effects”
Depth of field, cinematic_lighting:2, sharp_focus, film_grain, volumetric_lighting, bokeh, lens_flare, highres

This one is a doozy so hang tight.
Depth of field - It is basically how certain objects in the background / foreground are blurry.
cinematic_lighting - It adds more dramatic shading to the image.
Film grain - A slight imperfection in the fidelity of the picture, not every model is able to do that.
Bokeh, lens_flare - The funny blurry flying dots in the background.

Epic

What the hell is “Epic” style. It is a grandiose style made for scenes. It is over dramatic and how I create pictures like those two

1	2

The full prompt for those is very long and intricate and so we won’t be going over it but we will talk about the general setup and now hold your breaths:

Epic:2, Epic_scene:2, very_dramtic:2, Epic_theme:2, dramatic_scene:2

By itself you will get pointless slop.
Now modify it and add it not as a background preset but as an image prompt like this:
ADDBASE __MAIN_PERSONA__ (epic scene, epic, scene, Japanese theme)2 (epic)2 (best, masterpiece, no_lineart)+++,

This is a result of a very simplified prompt on Anime-Alpha
To explain it we basically tell AI we want an extremely generic picture with added some flavor on top.

---

This is basically everything you need to know about creating your own styles. Go and experiment with different prompts and loras!

This is an example of the same prompt as above but with a few loras and my own custom style preset.

---

54 tags style example: Courtesy of @leahbettsisdead

Physical	Clothes	Location	Prompt	Result
Horror And Macabre Elements, Atmospheric Perspective, Mythological And Barbarian Motifs, Flowing Drapery And Hair, Dramatic Chiaroscuro, Blended Realism And Exaggeration, Dramatic Use Of Contrast, Primal Energy And Savagery, Saturated Hues, Vibrant Color Palettes, Intense Emotional Expression, Romanticized Realism, Exaggerated Anatomy, Detailed Textures, Surreal And Dark Fantasy Themes, Foreshortening For Depth, Pulp Fiction Aesthetic, Epic Scale And Grandeur	Ray Tracing, Unreal Engine 5, Blender, Particle Effects, Chromatic Aberration, Volumetric Lighting, Global Illumination, Photorealistic Skin Textures, Glowing Wireframes, Surrealism, Octane Render, Metallic Shaders, Prismatic Refraction, 3D Render, Trending On Artstation	Pencil Sketch, Numerous Styles, Hand-Drawn, Messy Sketch, Ink Drawing, Sketch, Marker Art, Charcoal Sketch, Painted, Hentai Style, Anime Style, Scribbled, Inked, Black & White, Selective Color/Color Splash, Watercolor Art, Drawing	Speed Lines, Facial Blushes, Caricatures, Sparkling Effects, Sweat Drops, Detailed Costumes, Ecchi Fan Service, Pointy Chins, Symbolism In Eyes, Transformation Sequences, Cat-Like Mouths, Colorful Hair, Exaggerated Proportions, Large Expressive Eyes, Dynamic Hair Flow, Over-The-Top Poses	=====================

Physical	Clothes	Location	Prompt	Result
Horror And Macabre Elements, Atmospheric Perspective, Mythological And Barbarian Motifs, Flowing Drapery And Hair, Dramatic Chiaroscuro, Blended Realism And Exaggeration, Dramatic Use Of Contrast, Primal Energy And Savagery, Saturated Hues, Vibrant Color Palettes, Intense Emotional Expression, Romanticized Realism, Exaggerated Anatomy, Detailed Textures, Surreal And Dark Fantasy Themes, Foreshortening For Depth, Pulp Fiction Aesthetic, Epic Scale And Grandeur	Ray Tracing, Unreal Engine 5, Blender, Particle Effects, Chromatic Aberration, Volumetric Lighting, Global Illumination, Photorealistic Skin Textures, Glowing Wireframes, Surrealism, Octane Render, Metallic Shaders, Prismatic Refraction, 3D Render, Trending On Artstation	Pencil Sketch, Numerous Styles, Hand-Drawn, Messy Sketch, Ink Drawing, Sketch, Marker Art, Charcoal Sketch, Painted, Hentai Style, Anime Style, Scribbled, Inked, Black & White, Selective Color/Color Splash, Watercolor Art, Drawing	looking afar,high quality,8k,from below,low contrast,default,art bloom,anime image,glowing outline,masterpiece,good quality,high contrast,creative masterpiece,black background,dramatic churrascaro,best quality,1girl,light particles,chromatic aberration,flat color,neon yellow,upper body,flowing hair,silhouette,limited palette,(spot color),goat horns, (newest-)++ (absurdres-) (soft–) (hentai–) , (upright_straddle)+++ (kissing)+++, (silhouette)+++	===========================

HIGHLY IMPORTANT
When making your own custom styles remember to use not only sliders but also different presets both sfw and nsfw. They can drastically change the style of the picture

Example

#empty-sfw	#portrait-sfw	#empty-NSFW	#tentacles-nsfw

Above examples are only (Fully clothed)2

[Paul]

@Fuka

(“(Best quality) (perfect) (Highest quality)++ (Professional quality)++”,

Can you explain the reason behind using multiple, redundant quality expressions in a single prompt? Isn’t this just confusing the AI Image generator on what it is you’re looking for in the end?

[Fuka]

because they aren’t redundant. Due to how AI is trained different keywords have different datasets associated with them. Putting multiple quality keywords makes it so that only images that have all those keywords are used as a basis to create a new image.

In other words I’m using them to pinpoint a very specific artstyle / features when I’m doing that.

Hope this explanation helps

[Fortknox]

This is good stuff. A lot of the image tutorials seemed focused around anime. I’d love to see some realism prompts if there is major differences. I know realism images tend to be a lot stricter without as much flexibility as anime, so is there anything that helps with realism over anime in these prompts?

[Fuka]

Since the nature of realistic is more rigid when it comes to styles itself it might be hard to change things but you can actually change stuff when it comes to practical application. I will expand the answer to it in updated guide to styles.
Tldr: the basics of it still apply to realistic! You can easily change the artstyle with different prompts. Whether it be changing it to 3D render, more professional photography etc. Here is an example from my scribblefuka

Now we also can use loras which also change a lot:

Same character just with 3D LoRA
Here’s an example of my Nomillya

nsfw

Please do keep in mind that you need to apply more weight for it to work

Aside from that Leahs prompt also works with Realistic

Edit: If you want a help with how to approach making a style for realism. Think about it in more “professional” terms. bokeh, sepia, retro etc.

[Fortknox]

Thank you so much for the insight and speedy reply! Will definitely be spending the evening playing with all this!

[sunfanglei]

Your tutorial has been very helpful to me. I would like to ask you a question: how can we ensure that the generated images maintain a consistent appearance (look more like the same person, especially in terms of facial features, body shape, etc.)? The method I know is to modify based on a base image. I want to know if images automatically generated under the same kyn can maintain a consistent look or be as similar as possible？

[Fuka]

Sorry for late reply but here’s the deal, the image will have a consistent character only when you have a base physical prompt down, that means having your character be well described. This assures that when you’re playing with image prompts you will keep getting the same character.

---

It works like this:

The more vague your description is = more room for interpretation

The more precise it is = less room for interpretation

---

Please do keep in mind that you have to work within the prompts that the image model can understand. Anime models usually understand a lot from sites like danbooru etc when it comes to tags.

I would need to research about realistic because I don’t play with it. I will ask on discord whether or not smbd has found a smth like a taglist for realistic models.

---

[sunfanglei]

Thank you, I understand. This issue is a bit complicated, and I’m still trying.

[Fuka]

Here’s an example from Eralis

She tends to be mostly consistent aside from horns.

[Brimmstone]

Hey Fuka,
I was told by JetCity I should reach out to you. I’m super new to all these and I’ve been experimenting with all of this and reading as much as i can. I came across a Kyn today
https://soulkyn.com/l/en-US/@meridia to be exact. I’m struggling with figuring out how to make a Kyn look this good. Even in Picturize she looks amazing. I’m obviously missing something here. Any help would be greatly appreciated.