UnstableBastard ZIT 真人模型

UnstableBastard ZIT 真人模型是一款基于 ZImageTurbo 架构深度训练的高精度写实人像生成大模型，依托 S3‑DiT 单流扩散 Transformer 与 8 步极速蒸馏技术，在保持亚秒级出图效率的同时，大幅提升真人肤质、光影层次与面部细节还原度，同时适配亚洲人和外国人脸特征。

关于V3.5版本

V3 的重制版，加入了 capitan01R 提供的全新自定义节点（给他点个赞）+ 一些额外的 LoRA 节点

示例图像为 1024X1536，15 步，CFG 1.5（在 RTX3060 12GB 上，我大约需要 5-6 秒/帧，提示符执行时间大约为 100 秒 – 取决于提示符的长度）

• 就像 V3 一样，大多数采样器/调度器都能正常工作，而且彼此之间有很多不同之处，请测试并混合使用它们。有些支持主体合并，有些则不支持；有些适用于幻想场景，有些适用于“S”场景。尝试一下，尝试使用大于 1 的 CFG，尝试使用 10 步以上的设置等等。当我获得更多有趣的 LoRA 时，我会更新。
• 采样器：Euler_A / Euler
• 调度器：sgm_uniform / simple
• 步骤：8-15
• CFG 1 – 1.5
• 无光环流动偏移（更好的文本渲染，没有皮肤校正或其他偏移效果），
• 第一代产品如果使用这些设置，如果没有好的提示，有时会很糟糕，但第二代或第三代产品在不更改提示的情况下效果很好。
• 它应该可以与 ZIB 和一些 ZIT LoRA 在低强度下工作（未测试）。

作品参照

>A graffiti style artistic logo in the upper left corner of the screen that reads "Z-Bastard"<    A sultry, film-noir style portrait of a glamorous 1950s jazz lounge singer leaning on a grand piano, a lit cigarette between her lips sending wisps of smoke curling into the warm, golden pool of lamp light; dramatic chiaroscuro shadows, shallow depth of field as if shot on an 85 mm lens, rich vintage color grading with subtle film grain for a cinematic, high-resolution finish.
Negative prompt: blurry image , ugly , bad , Poor image quality, low image quality, dull colors, JPEG compression artifacts, low resolution,   censorship , censored , merged bodies ,  , bokeh , bad anatomy
Steps: 15, CFG scale: 1.5, Sampler: Euler a_sgm_uniform, Seed: 948267533630335, VAE: zImage_vae.safetensors, Size: 1024x1536, Model: Final V3.5.safetensors, Model hash: d2848d0630, Clip skip: 1

>A graffiti style artistic logo in the upper left corner of the screen that reads "Z-Bastard"<   **Subject**: A female figure with long, straight blonde hair (light golden hue). She is standing upright against a wall, facing the camera directly. Her right hand raises to gently touch her hair near the shoulder; left arm hangs relaxed by her side. **Quantity**: Single individual. **Action & State**: Static pose—body stationary, no movement implied. Facial expression neutral with direct eye contact. Skin appears smooth and well-lit (no visible blemishes or scars). The prompt does not require problem-solving; the image is a static portrait description of an individual in lingerie against a domestic interior setting. No need to devise alternative visual solutions—directly translating the scene into detailed instructions suffices. **Composition**: Subject occupies 70% of frame, positioned centrally with torso as focal point (chest and midsection emphasized). Left arm hangs vertically; right hand raised at shoulder height, fingers lightly grazing hair strands near ear level. **Setting**: Indoor domestic environment—plain white wall serves as primary background. To the subject’s left side (frame left), partial view of a closed door with no visible hinges or lock mechanisms. Right-side background features: - A white dresser/cabinet with multiple drawers, topped by a pink decorative item resembling floral-patterned fabric/box; adjacent to it rests black clothing/accessories on surface. - Above the cabinet hangs framed artwork displaying multicolored flowers (red/pink/blue petals) and text reading  Let's FUCK  in bold yellow letters. Floor visible at bottom edge of frame: gray carpet with no additional furniture or objects present beyond specified items. **Lighting**: Natural light source entering from left side—casts soft shadows on right-side wall (behind subject) and creates subtle highlights on hair strands, skin, and lingerie fabric. No harsh artificial lighting; ambient room illumination is dim but evenly distributed across visible surfaces. **Atmosphere**: Neutral indoor environment with no intentional mood-setting elements (e.g., candles, decorative lights). Focus remains solely on the subject’s physical appearance and attire details. - **Hair**: Light golden blonde color; long straight strands reaching below shoulders, parted down middle to frame face symmetrically. Texture smooth with no frizz or curls visible. - **Lingerie** (all black): - *Bra*: Mesh construction with lace detailing along edges and straps; central front clasp is metallic gold in shape of small square ring connecting two bra cups. Bra fits snugly, fully covering chest area without sagging. Straps attach to back via adjustable hooks not visible due to angle. - *Underwear* (thong/bodysuit): Mesh design with cutouts exposing pelvic region; waistband and straps feature repeated text “LOUNGE” in capital letters printed across fabric, separated by small metallic gold accents matching bra clasp. Straps wrap around hips symmetrically without visible buckles or ties exposed due to positioning. - **Tattoos**: Left arm (viewer’s left) displays two tattoos: - Upper forearm: Abstract geometric pattern with dark lines and shading; lower part of tattoo partially obscured by hair but extends from elbow toward wrist. - Forearm closer to hand: Distinct black ink drawing resembling stylized animal head profile, detailed facial features including eye sockets and nose bridge; no additional text present in this area. - **Skin**: Light tan complexion with smooth texture; minimal visible pores or imperfections under direct light exposure on torso/face areas. **Textual Content in Background Items**: Framed artwork above dresser/cabinet contains visible text: - Text reads "Let's FUCK"; font is bold yellow with irregular letter spacing, surrounded by multicolored flower illustrations (red/pink/blue petals). No other legible text on background items. No clothing/accessories beyond specified lingerie and tattoos are present in the image; no jewelry or personal effects visible except for subject’s own body markings and attire. Frame edges cut off partial views of door/hair but do not obscure core elements (face, torso).
Negative prompt: blurry image , ugly , bad , Poor image quality, low image quality, dull colors, JPEG compression artifacts, low resolution,   censorship , censored , merged bodies ,  , bokeh , bad anatomy
Steps: 11, CFG scale: 1, Sampler: Euler a_sgm_uniform, Seed: 796185385018290, VAE: zImage_vae.safetensors, Size: 1024x1536, Model: Final V3.5.safetensors, Model hash: d2848d0630, Clip skip: 1

图片提示词

>A graffiti style artistic logo in the upper left corner of the screen that reads "Z-Bastard V3"   A girl of Scandinavian descent with a tall, athletic frame, porcelain skin, and a light dusting of freckles across her shoulders; her platinum blonde hair is a messy, "just-woke-up" mane with effortless texture. She wears a tiny, ribbed grey crop top that fits snugly, paired with low-slung, oversized white lounge pants that rest on her hips to highlight her toned midriff. Her pose is playful and lithe; she is kneeling on a plush sheepskin rug, looking at the viewer a bright, mischievous gaze and a knowing smirk. A sun-drenched industrial loft with floor-to-ceiling windows, scattered art canvases, and a warm wooden floor. Lifestyle photography, shot on a Canon EOS R5, soft-focus background with high-definition skin textures and natural tones. Bright, high-key morning sunlight streaming through the glass, creating a soft halo effect around her hair and a radiant, glowing skin tone.
Negative prompt: blurry image , ugly , bad , Poor image quality, low image quality, dull colors, JPEG compression artifacts, low resolution,   censorship , censored , merged bodies , bad spatial awareness , bokeh , bad anatomy
Steps: 11, CFG scale: 1, Sampler: DDIM_ddim_uniform, Seed: 90604540525009, VAE: zImage_vae.safetensors, Size: 1280x1920, Model: Z-Bastard V3.safetensors, Model hash: f13faf5575, Clip skip: 1

(very pale skin, freckles), wet hair, two 20yo girls,  (one is ginger short hair. Others is blonde hair,  twinbraids, braces), stunningly beautiful, tall model like sexy bodyfigure, (a premium colored one-piece swimsuit with high cut hips and low back from), curvy round hips,  climbing out of a swimming pool, viewed from front, (water droplets on skin), one hand on pool edge, looking back at viewer over shoulder, (curvy silhouette, narrow waist), evening, poolside lights reflecting on wet concrete, neon signs from nearby bar in background bokeh, cinematic color grading, dramatic shadows, 50mm, film still, motion blur on water drips.
Negative prompt: blurry image , ugly , bad , Poor image quality, low image quality, dull colors, JPEG compression artifacts, low resolution,   censorship , censored , merged bodies , bad spatial awareness , bokeh , bad anatomy
Steps: 11, CFG scale: 1, Sampler: DDIM_ddim_uniform, Seed: 300875481437649, VAE: zImage_vae.safetensors, Size: 1280x1920, Model: Z-Bastard V3.safetensors, Model hash: f13faf5575, Clip skip: 1