Author's comment

As the title suggests, this article is for achievement purposes.

At first, I thought about making some kind of comic, and I even started three times, but then gave up... So I decided to just make a short picture article about the influence of artists and weights.

I don't think anyone will discover anything new, but maybe it will be useful to someone.

All images were regenerated on Civitai. I wanted to do this on socialdiff, but when I decided to start, there were problems with generation.

Important: You should understand that what happens next is a special case, not a something universal. Because how artist styles work and how weight is applied to them depends heavily on the models and loras you're using. Moreover, you most likely understand that this is not the only thing that affects artistic style)

Several images appear to be a generator error. They were generated several times, with the same result. Blame CivitAI for everything.

Some works and resources

Obviously, I'm not the first to write on this topic. So, you might want to check out other people's work.

The connection between iIllustrious and SDXL is clear. No further explanation is needed.

List of artists: https://huggingface.co/datasets/Laxhar/noob-wiki/tree/main. You're probably interested in danbooru, since the e621 is furry pit of hell (You may disagree with the author).

Viewer of artists' styles:

• https://www.downloadmost.com/NoobAI-XL/danbooru-artist/
• https://mzmaxam.github.io/Illustruous-Artists/ The images for this one had to be made on iLustMix9.
• https://rikkar69.github.io/SDXL-artist-study/ It's a mix of digital and traditional artists.
• • https://rikkar69.github.io/SDXL-artist-study/art-mediums/ An interesting addition (some mediums: Acrylic Paint, Polymer Clay, etc).
• • https://rikkar69.github.io/SDXL-artist-study/art-movements/ An interesting addition (some art movements: Art Deco, Art Nouveau, etc). You may also check this YouTube video and comment with art styles.
• https://sdxl.parrotzone.art/ I don't recommend it. It's a mix of digital and traditional artists. The loading times are insane.
• https://stable-diffusion.app/sdxl-artists/ As far as I remember, exclusively traditional artists.

Some Civitai.red article

• ARTISTS FOR ILLUSTRIOUS XL
• Yet Another Exploration of Illustrious Styles
• Yet Another Exploration of Illustrious Styles PART 2

Honestly, it's pretty good material, but if the images were grouped into a grid instead of being one big wall, it would be better.

Base image

Loading media image #1330...

Generation Mode: Create Image

Model: iLustMix (v11.1)

Lora:

• Detailer | Tool / Concept LoRA IllustriousXL:0.75,
• MoriiMee Gothic Niji | LoRA Style:0.5.

Why do I even use them? I almost always use this pattern with them, so it's just force of habit. I didn't add trigger words to the prompt. But it still doesn't matter, since the article about the special case.

Prompt: masterpiece, very awa, best quality, focused subject, dynamic angle, amazing quality, very aesthetic,{VAR1} detailed eyes, perfect eyes, realistic eyes, 1girl, solo, sitting, sitting in armchair, elegant dress, holding umbrella, umbrella blocking sun, long hair, beautiful face, lake background, forest background, shore, sunny day, sunbeams, peaceful mood, cinematic lighting, looking at viewer, movie perspective

Negaive Prompt: {VAR2} worst quality, low quality, normal quality, lowres, jpeg artifacts, watermark, signature, text, error, bad anatomy, deformed anatomy, pixels, missing limbs, missing arm, missing fingers, missing thumb, three fingers, extra limbs, extra fingers, imperfect eyes, imperfect hands, bad hands, distorted, ugly, two person, background characters

Generation Info:

• Size: 1024 x 1024
• CFG Scale: 4
• Sampler: Euler a
• Steps: 25
• Seed: 1392933500
• CLIP Skip: 2

As you understand, {VAR1} and {VAR2} in the base image are empty (default value). The article will be built around these two variables.

Usually it looks like: (by artist:weight) or (by artist_1:weight_1), (by artist_2:weight_2). For brevity, the structure section will contain abbreviations:

• 0.8 == ({VAR1} = (by artist_1:0.8),)
• 0.8/0.6 == ({VAR1} = (by artist_1:0.8), (by artist_2:0.6),)
• 0.8 == ({VAR2} = (by artist_1:0.8),)
• 0.8/0.6 == ({VAR1} = (by artist_1:0.8),; {VAR2} = (by artist_2:0.6),)

Structure

The rest of this article is devoted to applying artistic styles to the base image.

Selected artists:

• Afrobull - afrobull
• Aleriia_V (Lerapi) - aleriia_v
• Aroma Sensei - aroma_sensei
• BlushySpicy - blushyspicy
• Cutesexyrobutts - cutesexyrobutts
• Dandonfuga - dandon_fuga
• Kittew - kittew
• Mixplin - mixplin
• Sakimichan - sakimichan
• Slugbox - slugbox

Why them? Of all the ones I use, I remembered them the fastest.

Next come the sections by artist, which will contain groups of images according to the following scheme:

I recommend opening the images in a new tab and viewing them there (for PC, just zoom on phone), as you won't see anything in the article itself.

Image_1

Base Image	1.0

Image_2 (fuck the md table)

0.5	0.8	1.0	1.2	1.5
0.5	0.8	1.0	1.2	1.5
-0.2	-0.1	0.0	0.1	0.2
-1.0	-0.5	2.0	2.5	5.0

Image_3

For all artist:

1.0/1.0 artist_1/artist_2	1.0 artist_1	1.0/1.0 artist_1/artist_2	1.0/1.0 artist_2/artist_1	1.0 artist_2	1.0/1.0 artist_2/artist_1

Image_4

For all artist:

0.5/1.5 artist_2/artist_1	0.8/1.2 artist_2/artist_1	1.0/1.0 artist_1/artist_2	1.2/0.8 artist_1/artist_2	1.5/0.5 artist_1/artist_2

Why this set of weights? That's my decision.

As you can see, I didn't use negative weights in a negative prompt.

We could have increased the weight of artist_1 while keeping artist_2's weight fixed.

We could have used three artists...

There were many other possibilities, but this is just a short article with examples for getting the achievement!

Main part

Let's get started.

Afrobull

Loading media image #1331...

Loading media image #1342...

Loading media image #1352...

Loading media image #1362...

Aleriia_V (Lerapi)

Loading media image #1332...

Loading media image #1343...

Loading media image #1353...

Loading media image #1363...

Aroma Sensei

Loading media image #1333...

Loading media image #1344...

Loading media image #1354...

Loading media image #1364...

BlushySpicy

Loading media image #1334...

Loading media image #1345...

Loading media image #1355...

Loading media image #1365...

Cutesexyrobutts

Loading media image #1335...

Loading media image #1346...

Loading media image #1356...

Loading media image #1366...

Dandonfuga

Loading media image #1336...

Loading media image #1347...

Loading media image #1357...

Loading media image #1367...

Kittew

Loading media image #1337...

Loading media image #1348...

Loading media image #1358...

Loading media image #1368...

Mixplin

Loading media image #1338...

Loading media image #1349...

Loading media image #1359...

Loading media image #1369...

Sakimichan

Loading media image #1339...

Loading media image #1350...

Loading media image #1360...

Loading media image #1370...

Slugbox

Loading media image #1340...

Loading media image #1351...

Loading media image #1361...

Loading media image #1371...

Results

In all fairness, there could be some interesting conclusions here, but I suggest you draw them for yourself.

But we will of course say the simplest thing (everything in the context of this example):

• Should you write artists in the negative prompts?*
• • Probably not, unless you understand what exactly it affects.
• Should you play with the weights?
• • Of course, [-0.2, 1.5] are good weights, but you can go beyond those limits. Different artists react differently to this.
• Should you limit yourself to just these artists?
• • Of course not! If you already have a favorite model and an established pipeline, you can easily start searching for a combination of artists and weight that resonates with you.
• Do artists have a predisposition to certain colors and poses?
• • Of course they do, and this will be reflected in your generation unless otherwise specified in the prompt.

P.S. If you've read this far, please leave the names of artists you frequently use for generation in the comments. If they're different from mine, that's even cooler.

P.S.2. Drinking game: select a random image from this article. In 10 seconds, find all broken umbrellas. If you find all, no drinking; if you miss, drink by missing broken umbrellas.