造相 Z-Image-Base 全能文生图模型

Z-Image 是一款功能强大且高效的图像生成模型，拥有60 亿个参数。它目前有三个变体：

• 🚀 Z-Image-Turbo – Z-Image 的精简版，仅需8 次函数评估 (NFE)，即可达到甚至超越领先竞争对手的性能。它在企业级 H800 GPU 上可实现⚡️亚秒级推理延迟⚡️，并能轻松适配16G 显存的消费级设备。它在照片级图像生成、双语文本渲染（中英文）以及强大的指令执行能力方面表现卓越。
• 🧱 Z-Image-Base（此模型） ——未经精简的基础模型。通过发布此版本，我们旨在充分释放社区驱动的微调和自定义开发潜力。
• ✍️ Z-Image-Edit – Z-Image 的一个衍生版本，专为图像编辑任务而优化。它支持创意图像到图像的生成，并具备强大的指令跟随功能，允许根据自然语言提示进行精确编辑。

ComfyUI 原始模型：https://huggingface.co/Comfy-Org/z_image

原始 HF 仓库：https://huggingface.co/Tongyi-MAI/Z-Image

下载包含Text Encoder FP8/FP4和VAE

作品参照

A low-angle shot captures a sleek, tawny tabby cat mid-stride across a sun-dappled wooden floor, its body angled slightly left as it carries a small, weathered cardboard sign in its mouth — the word “Based” printed in bold black letters, slightly smudged at the edges from handling. The light streams through a high window behind, casting long diagonal shadows that stretch toward the viewer and frame the cat’s silhouette against a blurred background of stacked books and a half-open door. Dust motes dance in the beams, catching highlights as they drift past the cat’s fur, which glows with warm amber tones under the directional glow. Its tail curls low behind, barely visible but taut with motion, while its ears are pricked forward, focused on something unseen just beyond the frame. The sign tilts slightly downward, held loosely between its teeth, as if it’s been carried for some time — a quiet burden in this still moment of cinematic pause. Behind it, a dark armchair sits empty, draped with a rumpled wool blanket, and a single bare bulb hangs overhead, casting soft vignette shadows that deepen the mood. The shallow depth-of-field renders the background hazy, blurring the edges of furniture while keeping the cat’s eyes — sharp, intelligent, slightly narrowed — in crisp focus, drawing you into its determined gaze as it moves toward the viewer with quiet purpose.
Steps: 20, CFG scale: 3, Sampler: DDIM, Seed: 767308778170688, VAE: zImage_vae.safetensors, Size: 512x768, Model: z_image_bf16.safetensors, Shift: 5.0, Denoise: 1.0, Model hash: 996a67d3ff, Weight dtype: default, CLIP_1 Model name: Qwen3-4B-abliterated-bf16, Metadata generator version: 1.3.0

Hyper-realistic close-up selfie of a young East Asian woman in a snowy pine forest. Skin displays extreme macro detail: visible micro-pores, fine vellus hair, and natural peach-toned blush. Eyes feature high-clarity brown irises with distinct radial patterns and individual mascara-coated lashes. She wears iridescent pink-tinted ski goggles; the wide-angle lens reflection captures a detailed snow-capped mountain peak under a magenta alpenglow sunset. Wardrobe includes a voluminous white faux-fur headband with individual fiber definition and a metallic silver quilted puffer jacket showing specular highlights and microscopic fabric grain. Lighting is bright, natural daylight with a 5500K color temperature, creating soft shadows. Camera: 35mm focal length, f/2.8 aperture, ISO 100, providing sharp central focus with subtle peripheral depth-of-field blur. Environment: out-of-focus snow-laden conifers and a crisp blue-white sky gradient.
Negative prompt: cartoon, 3d render, anime, illustration, lowres, (bad anatomy:1.2), (deformed hands:1.3), (extra fingers:1.3), (blurry:1.2), (distorted reflection:1.2), (over-saturated:1.1), plastic skin, airbrushed, (missing limbs:1.2), (low quality:1.4), (bad eyes:1.2), fused fingers, (cluttered background:1.1), (watermark:1.2), (signature:1.2), (text:1.2), (glitch:1.1), (aliasing:1.1), (flat lighting:1.1), (asymmetric goggles:1.2)
Steps: 30, CFG scale: 5.5, Sampler: Euler, Seed: 527356789467977, width: 832, height: 1216

【设定】白天户外泳池中的定格瞬间，画面主体是一位中国小红书风格的年轻女性，大眼睛，可爱纯欲妆容，通体雪白，极致冷白皮，站立在一只粉紫色充气副浮具上，浮具漂浮于清澈的蓝绿色泳池水面，阳光充足，整体为真实生活摄影质感与纯欲取向审美结合的画面，纵向构图，人物处于画面中景略偏前的位置，成为绝对视觉中心。

【发型设计】深棕偏黑的中长发被水打湿后自然垂落，发根贴合头皮，发丝分束清晰，部分湿发贴在脸侧与锁骨位置，发尾有细小水滴，在日光下呈现深浅不一的自然反射，无刻意造型痕迹。

【OOTD全解】身穿白色连体泳装，材质为钩织针织面料，织纹细密规则，孔洞大小均匀，面料柔软但不透明；胸部结构简洁，贴合平胸轮廓，无明显支撑；侧腰大面积镂空，以同材质细绳连接固定，绳结与缝线清晰；下身剪裁稳定，包覆胯部。人物为明显梨型身材，上半身纤细，手臂瘦弱，肩背线条轻薄，胸部平坦，胯部与大腿比例突出，大腿超级粗壮，大腿根部屁股处有小巧黑色艺术字”REDZ“纹身，屁股超级大。右手腕佩戴发圈，手拿一杯蓝色鸡尾酒。

【场景与构图】前景为泳池水面与充气浮具边缘，中景为人物与副浮具主体，背景为虚化的泳池边缘、设施与人影轮廓，背景色彩被压低，形成清晰空间层次，视觉引导由水面反光与浮具线条汇聚到人物身体与面部。

【表情管理】人物妆容精致，底妆清透服帖，肤色均匀偏暖，鼻梁与颧骨有细腻高光；眼妆自然放大，睫毛卷翘，眼神明亮；唇妆为水润淡粉色。面部表情带笑，嘴角上扬，眼睛略微眯起，呈现出轻松、亲和、可爱的纯欲气质。

【肢体状态与细节】人物单脚站立于充气副浮具中央，一只脚稳踩浮具表面，另一只脚微微抬起，膝盖自然弯曲；双手向身体两侧张开以保持平衡，手指舒展，肩部放松，身体略微摇晃但被定格在平衡临界点，姿态显得俏皮而可爱，腿部肌肉与水珠细节清晰可见。

【光影设计】主光源为自然日光，自画面左上方照射，光线柔和均匀，在面部、肩部与腿部形成干净高光，阴影浅而自然，水面反射在下半身形成跳跃光斑，整体色温偏暖。

【拍摄要点】采用中等焦段镜头，浅景深处理，准确对焦人物面部与身体，快门凝固浮具轻微晃动与抬脚动作，曝光略偏亮，色彩干净通透，呈现典型小红书纯欲风的真实泳池瞬间。

画面你下方一行小字： CFG=1 STEPS=15 REDZDistill.test DX1
Steps: 15, CFG scale: 1, Sampler: DDIM, Seed: 666666, width: 816, height: 1536

关联模型