NukaCola Fearless版本v1.0 (ID: 557266)

Cold Nuka Cola, Fearless edition.

Brought to you with permission from DxDesignia

https://civitai.com/user/Dxdesignia

is a carefully tested, block weight merged fusion of my ColaNukaCola model + Fearless model from DxDesignia

https://civitai.com/models/425396/fearless

the fearless model I noticed did amazing yet subtle blushing faces and expressions, so this merge is calibrated to capture that feature, while maintaining the qualities of the original.

as always my aim is to create models that can handle 2d loras and concepts and bring them into stunning realism with as few mutations as possible. Creating realistic visions of impossible things or favorite characters is my aim.

there are subtle differences brought by this merge on skin texture but it affects composition, so it works very well as a refiner also,

my personal preference for use, is to allow ColdNukaCola to start the basic form of the seed, then use refiner to switch to fearless at 30% or 60% through

I'm getting amazing results at cfg6 27steps + 20step hiresfix with 4x ultrasharp

Refiner on both passes at 30% finishing each process with this fearless merge

this is my base negative prompt additive from doing the -7cfg trick, really boosts the realism:

((2d, greyscale, stick-figure_clip-art, cartoon, anime, blurry_fuzzy, low_resolution:2)) I also add (men, masculine) when appropriate

this is the result of inputting a simple pos prompt (gorgeous_woman) maybe, whatever it is you are mostly generating. then setting in scripts to run the model at negative cfg, I choose -7. the result is a series of awful, no-good, very bad images. by finding what they all have in common, and putting that as a negative prompt in normal gens, you can push the model from the other side, toward your goal.

I can't recommend this enough for getting any model to pop, mine included. it directly pushes your model away from it's own peculiar noise and worst tendancies, and is the easiest single way to have simple negative prompts that actually do something. it's just a free 15% boost to detail and quality.

use responsibly

I do not consent to having this model or it's demonstration images scraped and posted on other websites without permission (looking directly at you: Beone21 on tensor art, you rascal)

all other permissions are the province of the original creators of the merged components, as I didn't create these tools, simply mixed them as artfully as am able, to bring out their best qualities in pursuit of the height of stable realism

描述:

训练词语: