所以这篇文章，我把这一套真正的元提示词分享给大家，并一起来拆解看看，关于“锁定人物→生成变装→维持一致→加入风格层” 这套逻辑，用可复用的提示词结构写清楚。
三种风格
风格一：异次元糖果怪兽涂鸦
适用于：街头风、Maximalism、异次元插画
风格图示例：
提示词原文：
Use the uploaded photo.
Do NOT alter the person’s real appearance — keep the person’s face, body, clothing, colors, and texture completely photorealistic.
Do NOT change the background perspective.
Do NOT turn the person into a drawing or illustration.
Add a dense, overloaded layer of pop-style illustrated “sweets monsters” and graphic decorations ONLY around the person (and on top of their clothing if needed), but never on their skin or face.
Illustrated elements:
- many colorful cartoon monsters with thick black outlines, flat colors, and cute-but-ugly expressions
- sweets-inspired monsters: bananas, cookies, strawberries, melting chocolate, lollipops, ice cream, oranges, cupcakes, donuts, candy pieces, soda bottles, etc.
- additional graphic shapes: stars, hearts, arrows, drips, splashes, zigzag lines, exclamation marks, motion lines, sparkles, bubbles, comic-style text shapes (but no real text)
Make the decoration very dense and “busy”:
- fill the space behind the person with overlapping sweets monsters and shapes
- add monsters peeking from behind the shoulders, around the bag, at the person’s feet, and near the head
- allow some monsters and shapes to overlap the clothing and accessories (shirt, shorts, bag, shoes), but keep the skin of the face, arms, and legs photorealistic and visible
- use multiple layers of illustrations in front of and behind the person to create depth
- add glowing outlines, small white dots, and speed lines around the person to emphasize energy
Color and style:
- use a vivid, neon-like color palette (hot pink, yellow, cyan, lime, orange, purple, turquoise)
- keep all illustrated elements flat and graphic with clean edges and bold outlines
- ensure shadows and overlap suggest interaction with the real person (e.g., slight shadows on clothing where monsters touch)
Overall goal:
Create a highly decorated, maximalist pop-art scene where the real person stands in the middle, surrounded and wrapped by a chaotic crowd of playful sweets monsters and graphic doodles, while the person remains clearly photorealistic.
风格二：波普艺术插画风格服装
适用于：插画风格服装、人物服装插画
风格图示例：
提示词原文：
You are editing the uploaded real photograph.
Keep the person’s body, face, hair, skin texture, and the background exactly as they are.
Do NOT alter the person’s identity or any real-life details.
Transform ONLY the clothing into a clean, bold pop-art style illustration, while preserving the original outfit’s:
- shape
- silhouette
- color scheme
- pattern placement
- pocket positions
- neckline and seams
- major visual design elements
Illustration requirements:
- Use thick, smooth black outlines around the edges of the clothing.
- Convert folds and wrinkles into simplified cartoon-style linework.
- Preserve the original color palette, but make it brighter, cleaner, and more graphic.
- Convert textures, logos, or patterns into flat, vector-like versions.
- Add small stylish accents such as white highlight strokes or minimal motion lines around the clothing for a pop-art effect.
- Make the illustrated clothing look intentionally “drawn on top” of the real person, with crisp edges and solid colors.
- Ensure the illustration wraps naturally around the body as real fabric would.
Restrictions:
- Do NOT add any characters, animals, creatures, or mascots.
- Do NOT introduce any copyrighted elements.
- Only convert the clothing itself into stylized illustration; keep everything else fully realistic.
Goal:
Create a hybrid artwork where the person remains photorealistic, but the clothing is redrawn as a vibrant, vector-style pop-art illustration based on the original outfit’s design.
风格三：云隙白线巨型生物
适用于：梦幻感、光线极强、白线插画
风格图示例：
提示词原文：
Use the uploaded city photo as the base image.
Do NOT modify the real buildings, streets, people, or cars in the photograph. Keep the original scene fully realistic.
Add a very large illustrated creature in the sky, but draw it ONLY as clean, white line art — as if it were drawn with a white gel pen directly onto the sky. The creature should appear behind the clouds, partially visible through the natural gaps in the cloud layer.
Design of the creature:
- whimsical, gentle-looking fantasy monster
- drawn entirely with thin, glowing white outlines (no filled colors)
- composed of flowing shapes, fur lines, scales, floral curves, or surreal patterns
- only parts such as the face, eyes, horns, hands, or hair should peek through the clouds
- clouds must overlap the creature in some areas, making it feel truly hidden behind them
Integration with the sky:
- creature should softly glow white to contrast the blue sky
- maintain natural sky brightness and cloud structure
- use correct depth: buildings → clouds → creature behind clouds
- ensure some line art fades slightly where clouds are thicker, enhancing realism
Optional:
- add a few tiny white line-art elements in the sky (e.g., stars, floating shapes, small creatures) but keep them subtle and do not disturb the real scene.
Overall mood: a dreamy, surreal city scene where a giant, friendly white line-art creature quietly peers through openings in the clouds, seamlessly blending illustration with real photography.
也可以换成霓虹色线条
glowing neon colors outlines (for example soft neon green, neon yellow, and lime)
核心框架拆解
下面就是我对这一套结构化提示词框架进行拆解，
Imagine Prompt – 结构化提示词模板（锁定一致性必备）
—— 核心指令区（必填，不建议修改）
Based on the reference image, maintaining strict facial consistency and high likeness,
【填写核心角色特征，如：young woman with silver hair】
中文版：
基于参考图，保持严格面部一致性，高度相似度，
【填写角色核心特征，如：银发年轻女子】
这是锁脸的灵魂。 不要加太多形容词，不要改动结构，不要让模型跑偏。
——外貌与穿搭（变量区，可自由替换）
Wearing 【如：blue business suit】,
hairstyle: 【如：ponytail】,
detailed skin texture
这里是“变装区”，保持自由度，但不要动核心特征。
——动作与场景（变量区，可随意切换）
Action: 【如：drinking coffee】,
Location: 【如：in a sunny cafe】,
cinematic lighting, depth of field
你可以通过这里控制照片的风格演变，而不会破坏人脸一致性
——画质与负面词（固定区，不建议删改）
High quality, 8k resolution, masterpiece
--no distorted face, changing features, low quality, bad anatomy
中文版：
高质量，8K 分辨率，杰作
——没有面部畸形、五官变化、画质低、解剖结构糟糕
这是“模型纠偏区”，帮你避免 AI 乱改人脸。
使用说明（非常关键）
-
【】内容替换成你想要的内容即可
-
不要动“核心指令区”
-
想换造型？只改“外貌与穿搭”
-
想换场景？只改“动作与场景”
-
想保持长相永远不崩？保持结构不变
提示词工程的关键就是“结构化 + 分区化控制”。
为什么你做不到人像一致？
很多人以为一致性是靠“加参考图 + 多写几句保持一致”解决的。 但在提示词工程视角下，它其实包含 3 个维度的锁定：
-
身份特征锁定（Identity Features） 面部骨骼、五官比例、皮肤纹理，这是最基础的。
-
角色标签锁定（Character Tokens） 模型不是靠脸记人，它靠“特征组合”记人——比如： young woman, silver hair, soft jawline, wide-set eyes...
-
结构化提示词框架（Prompt Framework） 必须把“变的是变量、不变的是核心”写得足够清晰。
只有做到 分层、分区、分变量控制，AI 才不会崩脸。
提示词工程真正的思路
把这篇内容写完我更确定一件事情：
提示词不是越长越好，而是越“结构化”越强大。
优秀的提示词工程师，永远关注三件事：
-
固定关键区（保持一致性）
-
开放变量区（让创作自由）
-
风格层可堆叠（模块化组合）
如果你以后想做一整套自己的“角色宇宙世界观”，
如何使用？
昨天苏米分享了一个免费使用Nano Banana Pro的网站，无限制使用最新的Nano Banana Pro（昨天分享完今天就提示注册了，可以简单输入账密注册，没有限制）
用法很简单，就是上传图片+上面的提示词就能直接创作了：
声明：本站原创文章文字版权归本站所有，转载务必注明作者和出处；本站转载文章仅仅代表原作者观点，不代表本站立场，图文版权归原作者所有。如有侵权，请联系我们删除。
未经允许不得转载：异次元女孩风格提示词拆解，教你构建一套可复用的 AI人像一致性提示词系统