[maybe this model is best real texure in my models.]
Merge integration based on 3moon cartoon created a model with various textures and details. It responds well even when the small pixel size and CFG value are not high, saving details. However, personally, you should be careful of the burning phenomenon of the results.
The prom I used is like this.
[(ulzzang1.1:0.1), (pure eros:0.1):-0.05], 1girl, (semi realistic:1.3), (hyper realistic texture skin:1.3), korean mixed, kpop idol, blonde hair, blue eyes, blue hair, blue nails, blue scrunchie, bracelet, breasts, cleavage, collarbone, covered navel, earrings, eyeshadow, fingernails, fishnets, hand on hip, hands on hips, jewelry, large breasts, lips, looking at viewer, makeup, mole, mole on breast, multicolored hair, nail polish, pink hair, pink leotard, pink lips, pink nails, scrunchie, solo, twintails, wrist scrunchie, (teenager:1.3), (depth of field:1.3), (supreme intricate beautiful face:1.3), (supreme quality beautiful eyes:1.3), (supreme quality beautiful realistic skin:1.3) (hyper supreme quality beautiful women:1.1), (realistic lighting:1.2), (smooth lighting:1.2), (face beauty lighting:1.3), (erotic anime color:1.35)
Negative prompt: (lowest quality:1.3), (worst quality:1.3), (monochorme hair color:1.2), (black hair:1.1), (Femininity:1.35), (deteailed background:1.4), lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name, bad_prompt, small eyes, ugly, (sketch:1.25), (manga:1.25), (empty eyes:1.25), (fucked eyes:1.25), (Not REAL:1.25), (2D:1.25), (CG:1.25), (UNREAL:1.25), (anime:1.25), (black-white:1.25), (monochrome:1.25), (out of perspective:1.25), (flat perspective:1.25), [DeepNegative:0.2], uncomplete face, uncomplete eyes
Steps: 29, Sampler: DPM++ 2M Karras, CFG scale: 20, Seed: 3168580555, Size: 576x1024, Model hash: fc8b496104, Model: 3moon cartoon doll, Clip skip: 2, ENSD: -1, Dynamic thresholding enabled: True, Mimic scale: 20, Threshold percentile: 100, Mimic mode: Power Up, Mimic scale minimum: 10, CFG mode: Power Up, CFG scale minimum: 8, Power scheduler value: 6
[Ulzzang1.1 & pure eros & DeepNegative] is embedding.
i use dynamic threshold
Description
FAQ
Comments (9)
Can you add png with pnginfo please?
Couldn't get the dynamic thresholding to work on my Mac Studio so I couldn't reproduce the setup in the samples, but without it, this model generates images that could be pretty, but are noisy and blurry imo.
LOL.
Your sample images use Ulzzang1.1 & pure eros & DeepNegative embeddings?
I want to know what the model does BY ITSELF. That's how I judge whether or not it's a GOOD model.
ye. i use that embeddings. but don't use photo is well quality
@3moon please show an example of what the model can produce WITHOUT extra embeddings.
also need prompts!
So you want to know what Stable Diffusion does BY ITSELF and download custom models? Noone is forcing you 🤓
@omenizer I have no idea what you just wrote. Come back when you brush up on your English.
@wktra Maybe I am wrong, but embeddings don't alter the models capabilities, it more or less focuses them in a direction by training certain tokens, where as a LoRA will actually alter the model. You can think of it like this, you are the checkpoint, the prompt is the color pallette and the embeddings are the brushes. You can paint with your fingers, but if you use a paint brush you can add in the fine details and get exactly what you want. Either way, the result is your creation, it's just that one outcome used more tools while one did not. Though, I understand your sentiment, and having both with and without would be nice.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.