New Mecha illus 半写实模型

New Mecha illus 半写实大模型是一款基于Illustrious架构深度训练的AI绘画模型，兼具写实质感与艺术美感。它能精准捕捉人物神态、服饰纹理与场景光影细节，尤其擅长塑造风格鲜明的年轻女性形象，无论是可爱甜美的元气感，还是性感优雅的氛围感都能完美呈现，生成效果细腻自然，满足多样化的创作需求，支持nf/sw创作。

建议设置

我通常选择局部生成算法是Euler A or DPM++ 2M SDE Heun和SGM Uniform，根据我想要的细节程度选择20 或 40 步，根据我想要赋予模型的自由度选择CFG 4 或 7 。

展示图片中提供了两个网格用于演示。

你其实不需要“quality tags”，加你喜欢的就好。有时候，我只会加IllusP0s、SmoothPos或dfPos。我用这个模型 CLIP 做了一个“ MechaP0s ”（包含下载），因为它和传统的 Illustrious 风格略有不同。

对于否定提示，我通常会用IllusN3g（包含下载），不过，如果你想的话，也可以用你常用的否定提示工具 :D

Adetailer 也是个不错的选择，不过你喜欢用什么面部修复工具都可以。

禁止将此模型用于商业用途

以下是 sd-mecha 的配方：

import sd_mecha as sdm
sdm.set_log_level()

clip_uni = sdm.model("clip\\UniversalCLIPbase.safetensors", config="sdxl-sgm")
clip_v2 = sdm.model("clip\\illus2-clip.safetensors", config="sdxl-sgm")
recipe_clip = sdm.slerp(clip_uni, clip_v2)

recipe_base = (
  sdm.pick_component(sdm.model("illustrious\\pleasurechest_v1.safetensors"), "diffuser") |
  sdm.pick_component(sdm.model("vae\\sdxl_vae_extract.safetensors", config="sdxl-sgm"), "vae") | 
  recipe_clip
)

models = [
  sdm.model("hoj40c.safetensors"),
  sdm.model("shinv3.safetensors"),
  sdm.model("illustrious\\indexed_v1.safetensors"),
  sdm.model("illustrious\\artistrious_v10.safetensors"),
]

recipe_dare = sdm.ties_with_dare(recipe_base,*models,probability=0.6,seed=42,alpha=0.5)

lora = sdm.convert(sdm.model("..\\Lora\\000000\\HoJ4PC.safetensors"), recipe_dare)
recipe_lora = sdm.add_difference(recipe_dare, lora, alpha=0.3)

sdm.merge(recipe_lora, output="NewMecha.safetensors")

作品参照

1girl, solo, cowboy shot, Liadora, red hair, green eyes, black horns, three quarter view, looking away, ((latex g0thOutfit)), white lab coat, goggles around neck, contrapposto, legs appart, hand on own hip, dark cyberpunk city, rooftops, bright neon lights, cityscape, city lights, skyscrapers, skyline, night, overcast, rain, lightning, MechaP0s
 <lora:LiadoraOC:1>
Negative prompt: MechaN3g
Steps: 30, CFG scale: 4.2, Sampler: DPM++ 2M SDE Heun, Seed: 1785050630, TI: [object Object], Eta: 0.67, RNG: NV, Size: 1124x1636, Model: NewMechaExp, Version: f2.0.1v1.10.1-previous-665-gae278f79, Module 1: sdxl_vae_fp16_fix, Model hash: 8871470e87, Hires steps: 12, ADetailer VAE: None, Hires upscale: 1, Schedule type: SGM Uniform, Hires Module 1: Use same choices, Hires upscaler: None, ADetailer model: face_yolov8s.pt, ADetailer steps: 30, Hires CFG Scale: 4.2, ADetailer version: 25.3.0, Denoising strength: 0.3, ADetailer mask blur: 4, ADetailer confidence: 0.7, Diffusion in Low Bits: bnb-nf4 (fp16 LoRA), ADetailer dilate erode: 4, ADetailer inpaint padding: 32, Discard penultimate sigma: True, ADetailer use separate VAE: True, ADetailer denoising strength: 0.2, ADetailer use separate steps: True, ADetailer inpaint only masked: True, Clip skip: 2

MechaP0s
1girl, katsuragi \(senran kagura\), senran kagura, blonde,
big smile,
cute,  bottomless,
kneeling, in a giant present, torn gift wrapping,
holly, from above,
christmas, christmas decorations,
christmas sweater, dark, candle, wood,
comfy, warm, warm color, cozy,
Negative prompt: MechaN3g
Steps: 37, CFG scale: 3, Sampler: DPM++ 2M SDE Heun, Seed: 833232532, TI: [object Object], Size: 1248x1824, Model: IllustriousNewMecha_experimental, Version: f2.0.1v1.10.1-previous-669-gdfdcbab6, Model hash: 8871470e87, Schedule type: SGM Uniform, Denoising strength: 0.5, Clip skip: 2

图片提示词