Sign In

CMYK-CORE (Cyan,Magenta,Yellow)

39
233
39
13
Updated: Jul 28, 2024
styleneoncolorfulvividportraitkawaii
Verified:
SafeTensor
Type
LoRA
Stats
233
39
Reviews
Published
May 11, 2024
Base Model
SD 1.5
Training
Steps: 3,540
Epochs: 20
Usage Tips
Clip Skip: 2
Trigger Words
Cyan
Magenta
Yellow
Black
White
Hash
AutoV2
214B2ABF41
default creator card background decoration
Failed Generation Contest Participant
AI_AIDIA's Avatar
AI_AIDIA
Created on Civitai

First lora!! One fine day, my hands got itchy and tried out lora making for the first time using civitai lora training, I have to admit, I think this is a failed attempt from my intended result, or at least a 50% failure...

What does it do?
[STYLE+COLOR] It generates images with the CMYK color hue. It also accidentally creates a style that makes the character look more 2D with bigger and more colorful eyes, especially with close up.

What are the triggers?
The main color triggers are: Cyan, Magenta, Yellow, Black, White.
The colors can be customized by playing around with the combination, the position of the prompt affects the priority of the color, the color in front will be prioritized. Check my prompts for examples. If you use other colors, their hue may shift closer to CMYK.
Optional add-on prompts: ,1girl, solo, chibi, portrait, close-up, Simple Background, Pastel Color, Glowing, Neon
*There is a CMYK prompt BUT is not what you think...😢 I actually do not know what it does... There will be a little shadow cast on the face that's all. if anyone know, lmk.

Recommended Weight
Depending on the models used, the weight is good from 0.5 to 1.2.
On average, 0.6 to 1.0 have a more stable result.

Recommended model:
Any model will work well with the colors, but the style may not suit realistic model.
Overall it works best with anime/chibi models.
RE-LIVE_VIVIDPAST makes the colors POP if you want a burst of colors.

CORE Color Series:


As it was just a tryout, I just left most of the heavy lifting to Civitai, I auto tag everything, added my keywords, make sure my steps above 2,000 and poofs it's done. ☹️ I'll still commemorate my first lora by publishing it (the civitai recommended), hopefully I will find time to look through the other versions or revise and improve it. I know there are articles to guide me to make lora, but i got so loss all the time since i am not techy at all.

So just let me rant about parts that I felt is a failed attempt...😭😭😭

  1. "CMYK" Prompt - It isn't working the way I want it to, it got mistaken as a "Style" prompt rather than a "Color" prompt but when triggered, I don't even know what it does.

  2. Color biases - Although colors can be prompted individually, there is actually a bias in the order from Cyan, Black, Magenta, White, Yellow. It means cyan may appear when not prompt and yellow may not appear even when prompt.

  3. Shadowed face - Because of my data choice, the CMYK prompt will create a light shadow on the face, similar to a backlight effect.

  4. Model compatibility - Since I was just trying out, I chose SD 1.5 standard, while the color result works well with most models, some models will look brighter and others will look darker, there is also the issue with the style not working well on realistic models.

  5. Intended Effect - The intended effect is to create a full out effect of only CMYK hues that can lead to creative coloring such as pink skin etc., but the hue result came out more calm than expected.