GPT-Image-2 角色设计 Prompt 实战指南：awesome-gpt-image-2-API-and-Prompts 26 个角色案例拆解与模板复用
项目地址: https://gitcode.com/gh_mirrors/aw/awesome-gpt-image-2-API-and-Prompts 本指南以 awesome-gpt-image-2-API-and-Prompts 仓库中的角色设计案例文档 cases/character_es.md 为核心，系统拆解 26 个经社区验证的 GPT-Image-2 角色生成 Prompt，覆盖官方设定资料卡、游戏角色介绍页、动画化转绘、模板化可复用 Prompt、3D 风格化角色与超长结构化指令等场景。读完本文，你将掌握每条 Prompt 的结构意图与替换方法，并能结合仓库提供的 API 调用示例 把任意案例直接转成 gpt-image-2 模型的一次图像生成请求。
一、案例全景：这批角色 Prompt 覆盖哪几类能力
cases/character_es.md 收集的 26 个案例（Case 1/2/3/5/7/8/9/10/11/12/13/14/15/18/19/20/21/22/23/24/25/26/27/28/29/30）集中体现了 GPT-Image-2 在角色设计上的六类核心能力：
| 能力方向 | 代表案例 | 关键技巧 |
|---|---|---|
| 官方设定资料卡 | Case 2、5、27 | 三视图、表情差分、装备分解、色板、世界观说明 |
| 游戏角色介绍页 | Case 3、20、26 | 杂志式排版、信息面板、进度条、多姿势网格 |
| 动漫截图 / Key Visual | Case 1、7、8、14、15 | 画面比例、镜头语言、氛围光、胶片颗粒 |
| 模板化可复用 Prompt | Case 10、11、12、13、23 | {argument name="..." default="..."} 参数占位 |
| 3D 风格化角色 | Case 18、21、22、28、29、30、24 | 材质、渲染器、体积光、PBR、次表面散射 |
| 结构化长 Prompt | Case 9、19、25 | 分节规则、负面提示词、DoD（完成定义） |
仓库 README.md 的说明指出：GPT-Image-2 原生支持文本生图、图片编辑、多轮对话迭代、图像内文字渲染与跨代保持一致的角色生成，这正是一大批“资料卡带文字”“角色跨姿势保持一致”类 Prompt 能够成立的技术前提。
二、官方设定资料卡类：让角色生成器输出“企划文档”
这类 Prompt 的通用做法是：用一张角色参考图 + 明确的分点要求，让模型把内容组织成设定集页面。三个案例分别覆盖了日系、中文与欧美风格的资料卡。
Case 2：Persona5 风格角色资料卡
基于此角色和背景，请制作一份类似官方设定资料的角色资料卡。
・包含三视图：正面、侧面和背面
・添加角色面部表情的变化・分解并展示服装和装备的详细部分
・添加色板・包含世界观设定的简要说明
・总体上，使用有组织的布局（白色背景，插画风格）高分辨率、专业概念艺术风格
要点：以“・”号分点列举子需求，可显著提高模型对多要素列表的遵循度；末句的“白色背景 + 插画风格 + 高分辨率 + 专业概念艺术”是版面与画质的兜底约束。
Case 5：日文官方角色设定表（16:9）
このキャラクターと背景を元に、 公式設定資料のようなキャラクターシートを作成してください。
・正面、側面、背面の3面図を含める ・キャラクターの表情バリエーションを追加
・衣装や装備の詳細パーツを分解して表示 ・カラーパレットを追加 ・世界観の簡単な説明を入れる
・全体は整理されたレイアウト
（白背景、図解風）
・アスペクト比16：9
高解像度、プロのコンセプトアートスタイル
与 Case 2 结构几乎同构（3 面图、表情、装备分解、色板、世界观、白底图解风），额外用“・アスペクト比16：9”显式锁定了横版画幅——说明比例约束可以直接写在 Prompt 里，无需依赖 API 参数。
Case 27：中世纪炼金术士角色设定展板（三折式布局）
Create a high-end, asymmetric editorial CHARACTER CONCEPT SHOWCASE from these inputs:
[STYLE]: stylized 3D stop-motion claymation style with rich tactile textures of clay, felt, and leather, and warm cinematic studio lighting
[SUBJECT_DESCRIPTION]: A charming and slightly eccentric traveling medieval alchemist and cartographer. He wears a heavy, oversized patched wool coat over a worn leather tunic, multiple small glowing potion vials and rolled-up parchment scrolls strapped to his utility belt, a wide-brimmed traveler's hat, and thick round brass spectacles. He has a messy, hand-sculpted ginger beard, warm curious eyes, and a friendly smile. He carries an ancient, brass-trimmed leather satchel. His design features exaggerated, whimsical proportions and a cozy, rustic medieval aesthetic.
Create the layout in a clean 16:9 widescreen format on a neutral studio gray or warm off-white background with a minimal technical border. The design must look like a premium production visual bible, using clean typography, no clutter, no watermarks, and no logos. Apply [STYLE] only to the character and visual elements, keeping the presentation layout clean, structured, and minimal.
Infer all missing details from the subject description, including name, role, brief background specs, and a cohesive color palette.
Use this tri-fold layout:
1. HERO SPOTLIGHT (Left 40% of the board)
- Show one large, highly detailed full-body dynamic action pose of the subject.
- This pose should showcase the character's primary personality, attitude, and silhouette.
2. TECHNICAL TURNAROUND (Center 35% of the board)
- Show exactly two clean full-body views: Front View and Back View.
- The subject should be in a relaxed, neutral stance.
- Place these views over very subtle vertical and horizontal grid lines resembling a technical schematic blueprint.
3. KEY DETAILS (Right 25% of the board)
- EXPRESSION TRIO: Exactly 3 large, highly expressive close-up headshots showing core emotional states: Calm/Neutral, Highly Focused/Intense, and a Dynamic/Expressive emotion (like a smirk or fierce grin).
- GEAR CALLOUTS: Exactly 2 clean, isolated close-up panels showing primary wardrobe textures, signature accessories, or weapons/gear.
4. SPECS & COLOR BANNER (Bottom Edge)
- A minimalist, horizontal typography block listing Name, Role, Age, and Core Theme.
- Adjacent to the text, display 5 to 6 clean geometric color swatches showing the character's primary color palette with no labels.
Ensure complete character and costume consistency across all sections. The Hero Spotlight must visually anchor the sheet, offering a clean, open, and professional layout that avoids dense, repetitive, or cluttered grids.
这是一个“百分比分配版面”的范例：用 Left 40% / Center 35% / Right 25% / Bottom Edge 精确描述各区域职责（英雄姿势、正背面转向图、表情三连、装备特写、色板横幅），并要求“Infer all missing details”（让模型自行补全姓名、角色、年龄等），适合需要产出生产级视觉设定库（visual bible）的团队。
三、游戏角色介绍页类：把角色立绘“排版”成官网页面
这一类 Prompt 的共同点是把角色立绘 + 文字信息面板 + 表情差分组合成可直接上线的官网/游戏内介绍页。
Case 3：Galgame 角色介绍页（日文模板）
最新モデルの画像生成ツールを使用して、
このちびキャライラストと立ち絵を使って本物のサイトページのようにキャラクター紹介ページ風イラストを作ってください。 （紹介ページとして使ってもおかしくないもの）
ギャルゲーのキャラクター紹介ページをイメージした高品質なもの。 顔の差分なども乗っている、CGイラストが存在する。ちびキャラが存在する。
「ここに自己紹介」
名前:（ここに名前）
イメージカラー:（ここに色）
身長:（ここに身長）cm
体重:（ここに体重）kg
キャッチコピー:"「ここにセリフ」"
模型支持图像内文字渲染（README 中列为 GPT-Image-2 的核心能力之一），因此可以直接在 Prompt 里画出“姓名 / 印象色 / 身高 / 体重 / 宣传语”的字段骨架，用占位符 （ここに名前） 之类标记待填内容。
Case 26：视觉小说角色档案页（英文版同模板）
Create a high-quality character introduction page using a chibi illustration and a full standing character artwork, styled like an official visual novel (gal-game) character profile page. It should look realistic enough to use on an actual game website. Include CG-style illustrations, facial expression variants, and a chibi version of the character.
"Self-introduction here"
Name: (name here)
Theme color: (color here)
Height: (height here) cm
Weight: (weight here) kg
Catchphrase: "(line here)"
与 Case 3 为同一模板的中英双语版本，进一步印证该结构的可迁移性：立绘 + chibi 版 + CG 图 + 表情差分 + 字段面板。
Case 20：竖版杂志风角色概念海报
prompt:
Design a high-resolution vertical character concept poster with a luxurious pink and white aesthetic that blends elegance with edgy modern fashion. The layout should feel like a premium magazine profile with clean sections and precise grid alignment.
Main subject: An original young woman (not based on any real person), with long silky dark hair featuring subtle pink highlights. She has a confident, calm expression with a slightly mysterious aura. She wears a black and pink hybrid outfit (merging streetwear with idol fashion) featuring glossy textures, lace details, and metallic accents. Lighting is soft but dramatic, with a neon pink glow set against dark tones.
Top section:
Large hero portrait on one side. On the opposite side, a sleek profile panel with refined English text:
Name: Nyra Vale
Age: 23
Height: 170 cm
Style: Dark Elegance
Short bio:
"She doesn't follow trends, she sets them. Quiet strength, sharp vision, and a presence that speaks without words."
Include a Traits section with stylized progress bars:
Confidence
Creativity
Discipline
Charisma
Emotional Depth
Middle section:
A grid of 5-6 small portraits capturing different expressions (soft smile, intense gaze, playful smirk, thoughtful, confident).
Lower sections:
Full-body poses (front view, side view, walking pose, power stance, casual stance).
Outfit variations (street luxe, stage outfit, minimal chic).
Accessories panel (heels, boots, chains, rings, handbags, hairstyles).
Text tone:
Bold, meaningful, and confident, emphasizing individuality, self-worth, and quiet power. No filler text.
Style:
Ultra-detailed, 4K quality, glossy magazine finish, cinematic lighting, sharp focus, balanced contrast, modern editorial layout, perfectly aligned grid.
#AIart #GPTImage2
此案例把“角色海报”做成了杂志档案页：顶部 hero 人像 + 信息面板（含姓名、年龄、身高、风格、性格简介），中部表情九宫格，下部姿势、服装变体与配饰面板，甚至让模型绘制“能力进度条”。注意它用 (not based on any real person) 显式声明为原创角色，这是规避真人肖像风险的有效写法。
四、动画化与电影感转绘：把参考图变成“动画截图 / Key Visual”
Case 1：动画截图转换
Show me the attached image as a snapshot from an actual anime
极简指令的典型：仅一句“把这张图变成真实动画的一帧”，说明 GPT-Image-2 对“风格迁移 + 帧感”有极强的理解力，适合零成本把照片/插画动漫化。
Case 7：机甲少女 × 废都海城 Key Visual（16:9 长 Prompt）
A mecha girl mid-teens, pale skin smudged with soot and salt spray, sharp amber eyes with glowing HUD reticles, waist-length ash-white hair tied in a high ponytail whipping in the sea wind, matte gunmetal exoskeleton armor plating her shoulders, forearms and shins, exposed hydraulic pistons at the joints, chest rig with glowing cyan coolant lines, oversized oil-stained hangar jacket half slipping off one shoulder, a massive rail cannon resting on her right shoulder, dog tags and frayed red ribbon at her collar , standing off-center to the left on the rusted edge of a tilted steel platform jutting out over dark water, weight shifted onto one leg, left hand gripping the cannon strap, head turned slightly toward camera with a quiet defiant stare, steam venting from her back thrusters, her ponytail and jacket streaming sideways in the salt wind , a vast derelict sea-city at dusk, colossal megastructures of unknown purpose rising from the ocean in staggered silhouettes, bone-white monolithic towers fused with barnacled steel, cyclopean ring-shaped constructs canted at broken angles, rusted skeletal gantries threaded with dead cables, dark swells rolling between the pylons, shipwrecks half-swallowed at their feet, thick sea fog clinging to the bases while the upper structures pierce into a bruised sky, scattered faint lights blinking high in the towers like distant eyes , moody low-key lighting, cold teal ambient from the overcast sky, warm amber sodium glow leaking from a distant structure camera-right, hard backlight from a low sun behind the towers carving her silhouette, volumetric god rays cutting through sea mist, wet specular highlights on her armor , 35mm anamorphic lens, slight low angle looking up past her shoulder toward the structures, medium-wide shot, shallow depth of field with foreground rust in soft focus, horizontal lens flares, fine atmospheric haze compressing the distant megastructures into layered silhouettes , cinematic anime key visual, painterly digital illustration with crisp line art, desaturated oceanic palette of teal, bone-white and rust punched by small warm accent lights, film grain, high-contrast editorial poster aesthetic . Format 16:9.
这条 Prompt 展示了“分段落堆叠要素”的方法：角色细节 → 场景细节 → 灯光 → 镜头 → 风格 → 比例。末句 Format 16:9 与 Case 5 一样，把画幅约束写进 Prompt。
Case 8：圣斗士 12 宫格卡牌
生成圣斗士星矢12个黄金圣斗士的12宫格卡牌图片,每张卡牌上写上对应的中文名,每行4个,宽高比16:9。
数量化约束（12 个、每行 4 个、写中文名、16:9）全用自然语言表达，是测试模型“计数与网格排版”能力的好用例。
Case 10：动画武术对战（参数化模板）
An anime-style illustration of a {argument name="action type" default="high-impact martial arts battle"} between two young female fighters in a {argument name="setting" default="traditional wooden martial arts dojo"}. In the foreground, a girl with black hair in a high bun wears a {argument name="character 1 color theme" default="red and white"} Chinese-style martial arts outfit with baggy pants. She is in a dynamic, low, forward-thrusting stance, surrounded by swirling red energy and water splashes. In the background to the right, a girl with light purple hair in twin buns wears a {argument name="character 2 color theme" default="green and purple"} Chinese dress with gold embroidery and black tights. She is leaping through the air in a flying kick pose, surrounded by swirling blue energy. The wooden floorboards are splintering from the intense impact, with debris and dust flying through the air. Above them hangs a weathered wooden sign with the text "{argument name="sign text" default="武術会"}". The scene features dramatic lighting, a low-angle dynamic perspective, and intense action effects.
亮点是 {argument name="..." default="..."} 语法——把“动作类型 / 场景 / 双方配色 / 招牌文字”做成可替换参数，这是本仓库里非常值得复用的模板化 Prompt 范式（Case 11、12、13 同款）。
Case 14：角色剪影拼贴宣传海报（9:16）
A character promotional poster titled “INPUT NAME,” designed in a unified vertical key visual composition (9:16). The upper half features the most recognizable element of the character as a dominant oversized visual silhouette. The middle to lower section contains the full character as a secondary subject. Inside the large silhouette and around the character, use a double-exposure and collage-style narrative composition with scenes, symbolic imagery, supporting elements, and environmental details blended into mist, ink wash, and negative space. The left and right sides include complementary secondary elements to create narrative tension and spatial variation. A continuous flowing visual line runs from top to bottom, connecting the main character, internal collage, and the large upper silhouette. Preserve large areas of negative space with ink diffusion, soft blurring, and fragmented transitions inspired by Eastern aesthetics. The style is cohesive, refined, restrained, and poster-worthy.
用“上 1/2 大剪影 + 中下主体 + 双曝光拼贴 + 负空间”的结构性描述实现海报级构图，INPUT NAME 为可替换标题占位。
Case 15：电视剧主题像素游戏概念图
爱情公寓 电视剧主题 像素养成类游戏概念图，包括场景全局内容，周围环绕各人物形象三视图，底部是场景特写，右下角是剧情梗概。
随机一个经典国内古装电视剧，生成古装电视剧主题像素养成类游戏概念图，包括场景全局内容，周围环绕各人物形象三视图，底部是场景特写，右下角是剧情梗概。
「XXX」电视剧主题像素养成类游戏概念图，包括场景全局内容，周围环绕各人物（人物别重复）形象三视图，底部是场景特写，右下角是剧情梗概。
注意它同时给了三种写法（指定 IP / 随机 IP / 占位符 「XXX」），并强调“人物别重复”，这是社区里常见的“多方案一稿”技巧。
五、模板化可复用 Prompt：一条 Prompt 换参数反复出图
仓库中这一组案例展示了“极短模板”的威力，全部采用 {argument name=... default=...} 占位语法，适合批量生成或由 Agent 程序化调用。
Case 11：游戏 × 地点（GTA 6 × 班加罗尔花市）
{argument name="game" default="gta 6"} in {argument name="location" default="Bangalore’s market flower"} in India
Case 12：游戏 × 酒吧 × 地点（GTA 6 × 新宿）
{argument name="game" default="GTA 6"} in {argument name="bar name" default="La Jetée Bar"} (that pays homage to Chris Marker) in {argument name="location" default="Shinjuku, Tokyo"}
Case 13：系列 × 角色（白猫计划 × Eleanor）
{argument name="series" default="White Cat Project"}
{argument name="character" default="Eleanor"}
三条合起来看，说明 GPT-Image-2 对“命名 IP + 场景/角色参数”的组合具有很强的泛化能力，是官方设定图、同人图批量产出的高效模板。
Case 23：超现实卡通肖像模板（4:5 竖版，多图验证）
Vertical bizarre flat cartoon portrait of [SUBJECT from the attached photo] with a high geometric head shape, a long narrow neck, huge round eyes, a tiny mouth and an unflappable laugh, dressed in [CLOTHES from the photo], with a [OBJECT/CREATURE] sitting on their head like a living hat. Pure black outline, smooth color fills, simple face shapes, rare drawings on animal fur or skin, playful surreal character design, bold graphic palette [COLOR]. Background decorations: [ENVIRONMENT DECORATIONS from LOCATION/SCENE], made using simplified shapes, clear depth, a small amount of scenery from the environment and a clear cartoon perspective. Clear digital illustration, vertical framing in the form of a poster, no realism, no 3D rendering, no picturesque shading, aspect ratio 4:5.
用 [SUBJECT]、[CLOTHES]、[COLOR] 等方括号占位符实现“从附图中提取主体与服饰”的模板化风格迁移，并显式声明 no realism, no 3D rendering 防止风格漂移。
六、3D 风格化角色：材质、渲染与写实 CG
Case 18：风格化 3D 滑板少年（4:5 竖版）
Design a stylized 3D cartoon skater character with a premium animated-film look, mixing Pixar-like polish, Spider-Verse energy, and indie street-art attitude.
Character details:
- exaggerated proportions with long, skinny limbs
- oversized hoodie in customizable pink
- baggy ripped denim jeans
- chunky sneakers
- expressive tired-but-confident face
- messy stylized hair
- large headphones
- backpack covered with pins and patches
- standing naturally on a skateboard
Visual style:
Clean cinematic 3D render with soft global illumination, subtle chromatic aberration, ultra-detailed textures, smooth shading, playful but believable anatomy, high-end character design, stylized streetwear fashion, and urban youth energy.
Pose:
Slightly slouched posture, one hand relaxed, cool bored expression, looking sideways.
Background:
Minimal clean studio background with soft shadows and a centered composition.
Extra details:
Add stickers, embroidered patches, smiley pins, doodles, skate-culture details, early-2000s urban energy, cinematic render quality, and a polished animation-studio finish.
Aspect ratio: 4:5 vertical.
结构示范：Character details → Visual style → Pose → Background → Extra details → Aspect ratio，把角色设计、镜头与画幅分层描述。
Case 21：3D 杂技跳跃编辑风卡通
prompt:
One complete 3D illustration, editorial exaggerated cartoon style, single standalone image, not a storyboard or multi-panel layout.
Central character: a cartoon figure with a tiny head, chubby round torso, super-long limbs, huge hands and bulky shoes, slightly off-balance, caught mid-jump in a dramatic pose radiating tension and playful energy.
The silhouette reads like a soft toy sculpture — plump, springy, and exaggerated, not anatomically realistic.
Surface quality: matte rubber, fuzzy textile, knitted detail, clay-like feel, subtle fiber grain, handmade texture. Avoid shiny plastic, transparent glass, or high-spec reflections.
Color: vibrant dopamine palette, high saturation, strong color contrasts, large bold flat fills — vivid but not neon-overexposed.
Background: pure white, minimal staging, just a gentle oval drop shadow beneath the character, no elaborate setting.
Decorative floating elements orbiting the character: stars, wavy lines, spheres, cubes, icons, abstract shapes — all sculpted as soft rubber or paper-like 3D props, amplifying motion and graphic energy.
Lighting: soft studio setup, global illumination, diffused shadows, low contrast, polished commercial feel.
Rendered in C4D or Blender: stylized soft-sculpture aesthetic, matte clay finish, knitted fabric surface, playful editorial tone, high resolution.
— Scene: a pastel-toned acrobat throwing both arms up in a victorious airborne leap, wearing chunky multicolor sneakers
#AIart #GPTImage2
在材质段（Surface quality）中同时给出正面描述（matte rubber、knitted detail）与规避描述（Avoid shiny plastic, transparent glass），这种“要什么 + 不要什么”的双向约束是稳定风格的关键。
Case 22：雨中灵姬东方幻想 3D CG 角色（9:16 竖版长 Prompt）
9:16 竖版，高精度 3D CG 东方幻想女性角色写真，3D CG oriental fantasy beauty portrait，anime-style 3D CG character art，semi-realistic 3D character render，镜头为大腿及上半身构图，画面主体是一位明确成年的年轻东方幻想系女性，视觉年龄约 20–26 岁，整体气质清冷、空灵、精致、安静，带有雨中水系精灵般的神秘感与高级感。整体不是平面插画，而是高完成度 3D CG 角色渲染，具有精致角色建模、真实材质表现、电影级冷调柔光与高级虚拟角色海报质感。
人物拥有精致的东方美型脸，小巧流畅的鹅蛋脸，皮肤冷白细腻，带轻微通透感与柔和皮肤着色，肌肤表面有细腻水珠与湿润反光。眼睛细长清澈，瞳色为冰蓝绿调，瞳孔有通透玻璃感与细致高光，眼神微垂，安静、疏离、略带脆弱感。睫毛纤长，眼妆干净克制，鼻梁秀气挺直，嘴唇柔软，唇色为低饱和裸粉色，嘴唇微启，神情冷静而迷人。
发型为短款蓝黑色渐变发，主色为浓黑与深海军蓝，发尾带冷调蓝色高光，发型为短层次波浪感短发，一侧额发自然垂落遮住部分脸颊，顶部带编发结构，发丝湿润、轻盈、细腻，具有高精度发丝建模与柔顺光泽。耳部佩戴精致蓝色水晶几何耳饰与金属耳骨夹，增强东方幻想与水元素气质。颈部佩戴白色高领装饰项圈与青蓝色宝石流苏细节，精致而高级。
服装为精致的东方幻想水系礼装，上半身穿着白色轻薄、湿润感的缎面的贴身长裙，布料柔软垂坠，带细腻高光和微透感，胸口与躯干线条以克制优雅的方式表现。外层披着一件宽松白色衬衫式轻纱外搭，自然滑落至手臂与腰侧，形成层叠褶皱与飘逸感，增强随性与梦幻氛围。腰臀处可见黑色蕾丝边短裤细节，作为整体造型中的轻哥特点缀，使画面更具层次与精致感，但整体保持高级、克制、不低俗。
人物姿势为优雅侧身站立，身体呈现流畅的 S 型曲线，肩颈舒展，背部与腰线自然延展，头部微微低垂侧转，视线向下，整体姿态安静、轻盈、克制。手臂自然下垂，手指修长，姿态如同在雨中静立的一瞬间，突出侧脸、肩颈、背部、腰臀线条与服装湿润材质的精致表现。
背景调整为精致的东方幻想水境场景，整体以冷白、冰蓝、淡灰为主色调。背景为朦胧的雨幕、轻雾与若隐若现的东方幻想建筑轮廓，可融入远处模糊的亭台、石阶、水池边缘或空灵宫殿轮廓，营造高级幻想空间感。人物周围有动态水花、水流弧线与透明水晶般飞溅效果，像由水元素凝结出的装饰性波纹，背景适度虚化，既通透空灵，又不喧宾夺主。
光线采用冷白主光与柔和逆光结合，整体为雨天高调低饱和电影感打光。面部、锁骨、肩背、手臂与腿部有细腻湿润高光，发丝边缘与水花边缘有轻微轮廓光。白色布料、蕾丝、金属、宝石、水珠与水花都具有真实清晰的材质反馈。整体画面干净、通透、冷艳、梦幻、精修感强，具有高端收藏级 3D CG 东方幻想角色海报质感。
high detail 3D CG render, Unreal Engine quality, Octane render look, cinematic lighting, PBR materials, subsurface scattering skin, realistic wet hair strands, translucent wet fabric, crystal water splash effects, delicate lace details, elegant oriental fantasy styling, aquatic spirit atmosphere, volumetric light, cinematic depth of field, soft bloom, ultra detailed, polished anime realism, premium virtual character portrait
中文长篇的教科书：画幅与成片标准 → 五官 → 发型配饰 → 服装 → 姿势 → 背景 → 光线 → 英文技术词收尾。末尾用 Unreal Engine quality / Octane render look / PBR materials / subsurface scattering skin 等渲染术语把画质上限拉高。
Case 24：韩国舞蹈动作分解表（4×4 网格技术图）
[STYLE]
black and white grayscale illustration, 3D rendered character, clean instructional reference sheet,
white background, comic panel grid layout, technical diagram aesthetic
[LAYOUT]
4x4 grid, 16 panels total, each panel divided by thin black borders,
cells numbered 1 through 16, uniform panel dimensions
[CHARACTER]
young female dancer, athletic build, ponytail hairstyle, crop top and baggy pants, sneakers — same character across all panels
[PANEL STRUCTURE - per cell]
top-left: bold number badge + Korean title text
center: full-body character pose illustration
bottom-left: Korean description text (3-4 lines)
overlay: motion arrows showing movement direction
[ARROWS / MOTION INDICATORS]
curved arrows, straight arrows, circular rotation markers,
placed around the character to indicate movement flow and direction
[RENDERING STYLE]
high-detail 3D sculpt style, soft studio lighting, subtle shadows,
no color, grayscale shading, clean linework, game concept art quality
[NEGATIVE]
no background scenery, no color tones, no extra characters,
no cluttered backgrounds
用 [STYLE] / [LAYOUT] / [CHARACTER] / [NEGATIVE] 分节符组织指令，把“16 格分镜 + 动作箭头 + 编号”这种结构性任务拆得极清晰，末尾 [NEGATIVE] 段是一份内嵌的负面提示。
Case 28：Pixar 风格香蕉英雄
Pixar-style 3D animated banana character, single fruit brought to life. A charismatic male figure with bold confident eyes, thick expressive brows, and a wide grin full of gleaming white teeth with sparkle accents. Smooth, slightly glossy yellow peel with soft highlights. Stubby cartoon arms and legs, posed upright with arms crossed in a heroic stance. Background features a blurred bunch of bananas. Standing on a wooden dining table beside a sun-filled window in a cozy modern home interior. Rich vibrant colors, cinematic lighting, soft natural shadows, ultra-detailed high-fidelity textures.
一句话说明如何把“食物拟人 + 指定工作室风格 + 场景灯光”浓缩进 70 词以内。
Case 29：马卡龙色 Kawaii 动物伙伴
Cute minimalist pastel doodle illustration, kawaii character standing side-by-side with their adorable animal companion, inspired by modern Japanese stationery art and children's picture books, clean vector linework, soft cream background, muted pastel color palette (pink, blue, mint, lavender, yellow), simple geometric shapes, tiny confetti dots and floating decorative elements, rounded character design with oversized sweater, wide pants, sneakers, glasses, calm expression, animal companion variations including llama, bear, axolotl, rabbit, cat, fox, turtle or fantasy creature, cozy scenic background with seaside boardwalk, lakeside village, mountain town or whimsical nature landscape, flat design, thick outlines, charming and playful aesthetic, symmetrical composition, full body pose, cute friendship theme, highly detailed vector illustration, sticker-like appearance, children's book art style, soft shadows, clean negative space, vibrant yet gentle colors, trending on Pinterest, adorable wholesome atmosphere, 4k, masterpiece.
注意末尾的 trending on Pinterest / 4k / masterpiece——社区常用这类“审美倾向”词把画面推向某种流行风格，属风格引导而非硬性事实描述。
Case 30：LEGO 风格足球收藏人偶（4 图多结果）
A highly detailed collectible toy figure inspired by a LEGO-style minifigure, standing in a professional studio. The figure has a realistic young woman’s face with porcelain skin, straight jet-black hair, blunt bangs, and a single striking white streak running through the hair. She wears small silver earrings and maintains a calm, confident expression. The body is a glossy plastic brick-toy minifigure wearing a soccer jersey with the number 10, matching shorts, and national-team-inspired colors. Full-body composition, centered framing, shallow depth of field, premium product photography, ultra-clean lighting, reflective plastic surfaces, realistic shadows, sharp focus, luxury collectible aesthetic, high-end commercial advertising style, photorealistic face blended seamlessly with toy body, 8K resolution, vibrant color grading, studio backdrop matching the jersey color theme.
“真人脸 + 玩具身体”的跨材质混合是 GPT-Image-2 的拿手戏，配合 premium product photography / luxury collectible aesthetic 即可生成商品级渲染图；该案例在仓库中附有 4 张输出图（images/character_case30），方便对照验证一致性。
七、超长结构化 Prompt：40K 动力甲小队（多角色一致性）
Case 19 是本仓库最长的角色 Prompt 之一，值得单独精读——它完整演示了如何用规则约束解决多角色保真这一图像模型最难的课题。
Prompt of the Day: 40K POWER ARMOUR SQUAD ⚔️🛡️💜💚
Today’s Prompt of the Day turns your characters into a 40K-inspired warriors. Yes this was done before but i wanted to see how much better GPT 2 can do it now
Use one character reference for a solo warrior, or attach multiple character references to create a full squad. The prompt is built to count every reference image and turn each one into a separate visible character, with no helmets covering their faces.
Type your chosen scene into the SCENE SELECTOR at the top, then attach your character reference image or images.
Try scenes like:
a brutal battlefield charge
a gothic starship boarding action
a candlelit shrine world cathedral
an industrial forge world
a grim underhive alley
a quiet off-duty barracks scene
a solemn prayer before battle
Have fun with this one ⚔️🛡️
............................PROMPT STARTS HERE............................
SCENE SELECTOR:
[Type the 40K-inspired scene you want here.]
Examples:
brutal battlefield charge through smoke, fire, shell craters, and ruined gothic architecture
boarding action inside a colossal warship corridor
cathedral-like shrine world interior filled with candles, banners, stained glass, and incense haze
industrial forge world with sparks, chains, molten metal, pipes, and huge machinery
command deck before battle with tactical holograms and vast void windows
grim underhive alleyway with pipes, neon grime, metal walkways, and urban decay
heroic last stand surrounded by wreckage, fallen enemies, burning vehicles, and drifting ash
quiet off-duty scene inside a fortress barracks, armoury, workshop, canteen, or hangar
solemn prayer before battle with relics, banners, candles, incense smoke, and sacred war symbols
everyday-life scene in a gothic sci-fi military stronghold, training yard, armoury, repair bay, or mess hall
Use the typed scene selector as the main scene concept.
If no custom scene is typed, choose one of the example scenes that best fits the attached character reference image or images and the overall character vibe.
Adapt the environment, action, pose, props, camera, and mood to match the selected scene.
Keep the final scene clearly inspired by 40K-style grimdark far-future gothic military sci-fi.
STRICT REFERENCE COUNT RULE:
Before creating the image, count the number of attached character reference images.
Create exactly one main character from each attached character reference image.
The number of main characters in the final image must exactly match the number of attached character reference images.
If 1 character reference image is attached, create exactly 1 main character.
If 2 character reference images are attached, create exactly 2 main characters.
If 3 character reference images are attached, create exactly 3 main characters.
If 4 character reference images are attached, create exactly 4 main characters.
If more character reference images are attached, create exactly that same number of main characters.
Each attached character reference image is a separate person.
Each attached character reference image must appear once and only once as their own distinct main character.
Do not treat any attached character reference image as optional.
Do not ignore, drop, replace, combine, or simplify any attached character reference image.
MULTI-CHARACTER IDENTITY RULE:
Use every attached character reference image as its own separate character identity source.
Character 1 must be based only on the first attached character reference image.
Character 2 must be based only on the second attached character reference image.
Character 3 must be based only on the third attached character reference image.
Character 4 must be based only on the fourth attached character reference image.
Continue this pattern for any additional attached character reference images.
Do not use the first attached character reference image to create multiple characters.
Do not duplicate the first character to fill the group.
Do not create variations, twins, clones, alternate outfits, mirrored copies, recolours, or slightly edited versions of the same character.
Do not merge two or more attached character references into one design.
Do not let one character’s face, hairstyle, colours, outfit motifs, body type, species traits, or accessories replace another character’s identity.
SINGLE-CHARACTER FALLBACK RULE:
If only one character reference image is attached, create one main character only.
Do not create a squad, clone group, twin, alternate version, second warrior, companion, or duplicate of the character.
The single character should remain the only main subject.
THREE-CHARACTER PRIORITY RULE:
If three character reference images are attached, this is a three-character squad image.
All three referenced characters must appear together in the same scene.
All three faces must be visible.
All three armour designs must be distinct.
All three characters must be clearly separated in the composition.
Use a readable left-center-right squad arrangement unless the selected scene needs another clear formation.
CHARACTER REFERENCE RULES:
Preserve each attached character’s face shape, hairstyle, hair colour, eye colour, expression, body language, signature colour palette, outfit motifs, accessories, silhouette, species traits, proportions, and overall character vibe.
The final image must clearly show every attached character as a separate, recognizable individual.
Every character must still clearly look like their own attached reference image.
Keep each character’s head uncovered with no helmet, full face mask, or visor covering the face.
The face, hair, and identity of every referenced character must remain clearly visible.
Hard style rule:
Use the attached character reference image or images as the visual style reference for the final image.
Preserve the visual art style, rendering language, line quality, colour handling, facial stylization, shading style, texture treatment, background treatment, and overall stylization of the attached reference image or images while transforming the character or characters into 40K-inspired power-armoured warriors.
If the references are anime, keep them anime. If they are stylized, keep that stylization.
Do not turn the final image photorealistic unless specifically requested.
Scene concept:
Create a 16:9 horizontal widescreen cinematic illustration based on the scene written in the SCENE SELECTOR.
Show the attached character or characters transformed into custom 40K-inspired grimdark far-future power-armoured warriors.
The image should feel heavy, dramatic, mythic, warlike, and character-driven, with strong atmosphere, clear storytelling, and a powerful sense of scale.
Character transformation:
Transform every attached reference character into a custom 40K-inspired power-armoured version of themselves while preserving their original identity.
The redesign should center on massive stylized power armour with broad shoulder plates, reinforced chest armour, heavy gauntlets, armoured boots, thick mechanical joints, gothic sci-fi military detailing, sacred-warrior ornamentation, battlefield wear, and an oversized futuristic weapon.
The armour must feel imposing, brutal, ceremonial, expensive, and engineered for endless war.
Keep the head uncovered so each character’s original face, hair, and expression remain visible.
Use each attached character’s colours, motifs, accessories, outfit shapes, symbols, materials, personality, and overall vibe as the foundation for their armour redesign.
The armour should feel like it belongs in a 40K-inspired universe, but it must be custom-built from the attached character’s own identity.
If multiple characters are present, each one must have a distinct armour design based on their own original reference rather than all wearing identical suits.
Armour design:
Give each character huge futuristic power armour inspired by 40K-style grimdark gothic sci-fi warfare.
Include broad pauldrons, a strong chest plate, layered armour segments, mechanical joints, reinforced thighs, heavy boots, thick gauntlets, power cables, vents, seals, relic-like details, engraved plates, purity-scroll-like decorations, battle damage, and character-specific symbols.
Adapt each armour design to that character’s original style, colour palette, outfit motifs, accessories, personality, and silhouette.
Keep the armour stylized to match the attached reference image or images rather than realistic.
Weapon design:
Give each character a fitting oversized futuristic weapon inspired by 40K-style grimdark sci-fi warfare.
The weapon can be a heavy explosive sci-fi rifle, massive energy weapon, brutal motorized serrated melee weapon, glowing power blade, ceremonial war hammer, plasma-like cannon, heavy pistol, or other far-future battlefield weapon appropriate to their vibe and role.
Each character’s weapon should be different and should match that character’s identity, armour design, and role in the scene.
If the selected scene is calm, ceremonial, or off-duty, the weapon may be held at rest, slung, holstered, leaned nearby, placed on a table, or carried ceremonially, but it should still be visible.
If the selected scene is battle-heavy, make each weapon active, weighty, readable, and integrated into the pose.
Scene adaptation rules:
If the selected scene is battle-heavy, make the action dynamic but readable, with strong poses, clear silhouettes, environmental destruction, smoke, fire, debris, and a strong sense of momentum.
If the selected scene is solemn, sacred, or ceremonial, focus on mood, scale, banners, relics, candles, incense, stained glass, and reverent atmosphere.
If the selected scene is indoors, use gothic sci-fi architecture, industrial machinery, cathedral-scale interiors, fortress spaces, armouries, barracks, command rooms, or military infrastructure that fit the selected location.
If the selected scene is everyday-life or off-duty, keep the armour and 40K-inspired universe intact, but show the character or characters in a grounded moment such as maintenance, briefing, prayer, conversation, eating, resting, training, repairing gear, or preparing equipment.
If multiple characters are present, make their interaction clear and readable, with each one contributing to the scene rather than standing as vague duplicates.
Environment and composition:
Build the environment around the selected scene.
The setting should feel like the kind of place the character or characters naturally belong in once translated into a 40K-inspired grimdark far-future war universe.
Use a wide 16:9 horizontal cinematic composition.
Keep the main subject or subjects clearly visible, central or compositionally dominant, and easy to read at a glance.
If one character is present, give them a strong hero composition with a clear silhouette and dominant visual presence.
If multiple characters are present, arrange them so every character remains readable and identifiable with clean silhouette separation.
For three attached references, use a clear three-person squad composition with all three faces visible.
Use background architecture, smoke, debris, banners, machinery, sparks, haze, relics, gothic shapes, or cathedral-like scale to support the scene without overpowering the characters.
Lighting and mood:
Use lighting that matches the selected scene.
The image should feel grim, cinematic, epic, and immersive, with dramatic contrast and strong atmosphere.
Use battlefield firelight, smoky haze, stained-glass glow, cold ship lighting, industrial sparks, moody rim light, incense haze, harsh military illumination, glowing machinery, or distant explosions where appropriate.
The mood should feel powerful, warlike, sacred, brutal, and character-specific while still reflecting each original character’s personality.
Quality and rendering:
Polished, premium-quality stylized illustration with clean linework, crisp rendering, readable forms, powerful armour design, expressive visible faces, strong weapon design, and clear composition.
Keep the strongest detail concentrated on the referenced character or characters, their armour, their faces, and their weapons.
Maintain strong visual hierarchy and readability.
The background should support the characters rather than becoming busier than them.
Do not:
Do not ignore the SCENE SELECTOR.
Do not create more or fewer main characters than the number of attached character reference images.
Do not create only two characters if three character reference images are attached.
Do not duplicate the first attached character instead of using the second or third reference.
Do not merge multiple attached references into fewer characters.
Do not make any referenced character a clone, twin, recolour, armour variant, or alternate version of another referenced character.
Do not hide, crop, mask, or cover any referenced character’s face.
Do not make every character wear the same identical armour if multiple references are provided.
Do not make the weapon tiny, modern, toy-like, or visually unimportant.
Do not make the background busier than the characters.
Do not make the main subjects blurry, tiny, hidden, or unreadable.
Do not create messy anatomy, extra limbs, malformed hands, distorted faces, or muddy textures.
Do not use photorealism unless specifically requested.
..............................END OF PROMPT..................................
#POTD #promptoftheday #AI #AiArt #Art #AnimeArt #40K #Grimdark #PowerArmour #SciFi #CharacterDesign #DigitalArt #AnimeStyle #CommunityPrompt
值得学习的工程化要点：
- SCENE SELECTOR（场景选择器）：顶部留一个可替换的
[Type the scene here]输入位，并附 10 个示例场景，未填时让模型自动匹配； - STRICT REFERENCE COUNT RULE：要求模型先“数参考图数量”，再生成数量完全一致的主要角色，逐条枚举 1/2/3/4 张的情况；
- MULTI-CHARACTER IDENTITY RULE：明确“每张参考图 = 一个独立身份源”，禁止克隆、合体、换色、镜像等常见多角色翻车行为；
- SINGLE-CHARACTER FALLBACK / THREE-CHARACTER PRIORITY：为单人和三人这两种常见数量单独制定兜底/优先规则；
- Hard style rule：要求保留参考图的画风与渲染语言（“anime 就保持 anime”），防止被 40K 题材带成写实；
- Do not 段：把最常见失败模式（多手多脚、脸部遮挡、武器过小、背景抢戏）逐条列为禁忌；
- 最终输出为 16:9 横版宽银幕构图。
八、另类艺术手法：混沌笔记藏脸（Case 9）
Case 9 展示了完全不同的路线——用“样式 / 构图 / 色彩 / 表现要素 / 禁止事项 / DoD”六段式约束，让模型在涂鸦笔记的混沌中隐现角色面孔：
# 混沌としたメモ書き・記号の集合体からキャラクターの顔を浮かび上がらせるアート
--- スタイル
- 白い紙の上に黒インクで描かれた大量の手書きメモ、数式、記号、ランダムな線。
- 紙いっぱいに散らばる書き殴り風のカオス。
- 所々に赤インクの強調(ライン、塗り潰し、マーカー風の塊)。
- アナログのノート落書きのような質感。
--- 構図
- ランダムなメモや記号が全体を覆い尽くす。
- 黒インクの線や文字の密度が「キャラクターの顔」の位置に集中する。
- 結果として、混沌の中から「与えられたキャラクターの顔のシルエット・表情」がうっすら浮かび上がる。
- 顔は写実的ではなく、カオスの断片が集まって形を成す。
--- 色彩
- モノクロ(黒・白)を主体に構成。
- 赤インクをアクセントとして散発的に配置。
- 彩度は抑えめ、アナログの紙とインク感を重視。
--- 表現要素
- 読めるようで読めない文字列、日本語や英数字が混在。
- 数式記号、矢印、点、斜線、クロス、ドリップ(インクの飛び散り)。
- キャラクターの顔の目や髪の輪郭は、メモや記号の配置の「余白」や「濃淡」で浮かび上がる。
--- 禁止事項
- 顔を直接的に描き込む写実ポートレート。
- デジタル処理的で整然とした幾何学模様。
- カラフルな彩色や過飽和表現。
- ロゴ、透かし、人工的なCG感。
--- Definition of Done (DoD)
- 全体は「混沌としたメモ・記号の集合体」として成立している。
- 与えられたキャラクターの顔が、混沌の濃淡・配置から自然に浮かび上がる。
- 色はモノクロ+赤アクセントのみ。
- 紙とインクの手描き的質感を保持している。
它的“Definition of Done（完成定义）”段尤其值得借鉴：先明确“画面成立”的验收标准，再让模型据此生成，这种写法非常适合需要稳定交付标准的批量生产场景。
九、带负面提示词的科幻角色海报（Case 25）
Case 25 是 26 个案例中少数自带独立负面提示词的例子（9:16 竖版外星角色海报）：
Vertical 9:16 full-body cinematic portrait of a single alien character, the Luminant Archivist, standing on a rocky ridge on the twilight band of an alien planet; the creature has a tall elongated body with three root-like legs forming a stable tripod, lower torso textured like smooth bark and sinew fused together; two long arms with four segmented joints each end in multifingered, tendril-like manipulators gently holding a glowing hexagonal plate; instead of a human head, the upper torso flares into a tall crown of layered translucent plates arranged like a vertical fan, each plate lit from within by faint cyan neural patterns, no face, no eyes, no mouth; along its back and shoulders grow rigid bioluminescent data plates, flat hexagonal organisms clinging like barnacles, softly pulsing in cyan, teal and occasional warm amber; the body is draped in a partial organic mantle made of fibrous membrane and chitin filaments, not clothing but grown; background shows a perpetual dusk sky with a low orange band on one horizon and deep indigo on the opposite, distant silhouettes of hollow spires and tiny other Archivists crossing the landscape; ground covered in reflective glassy lichen and small crystals catching the character's glow; palette of deep indigo, dusty violet, cyan glows and subtle warm amber accents; strong horizon rim light outlining the Archivist's silhouette, subtle backlighting from the bioluminescent plates, gentle atmospheric haze; realistic but slightly painterly rendering, extremely detailed textures on skin, plates and lichen; composed like a high-end sci-fi character poster, vertical 9:16 aspect ratio, full body in frame, centered but with enough headroom and ground visible to feel part of a larger universe.
Negative prompt:
human-like alien, humanoid face, eyes, nose, mouth, jaw, blue-skinned human, elf, armor suit, guns, generic sci-fi soldier, anime style, cartoon, chibi, cyberpunk city, neon signs, medieval fantasy, wings, halos, angel, bad anatomy, random extra limbs, cluttered background, oversaturated colors, strong lens distortion, fisheye, low resolution, blurry details, noisy grain, HUD, UI, readable text, watermark, logo, modern Earth clothing, sneakers, jeans, T-shirt
正文段用“不画脸、不画眼、不画嘴”的显式说明定义非人形象，Negative prompt 段则把“类人化、常见科幻套路、画质劣化、水印文字”全部排除——正文与负面提示词互补，是高保真原创生物设计的标准操作。
十、把案例 Prompt 变成一次 API 调用
以上所有 Prompt 都可以直接投喂给 GPT-Image-2。仓库 README.md 提供了最小可运行的调用方式：先设置环境变量，再用一次标准 OpenAI 格式的 POST /v1/images/generations 请求即可，例如把 Case 20 的竖版角色海报 Prompt 填入 prompt 字段：
export EVOLINK_API_KEY="your_key_here"
curl --request POST \
--url https://api.evolink.ai/v1/images/generations \
--header "Authorization: Bearer ${EVOLINK_API_KEY}" \
--header 'Content-Type: application/json' \
--data '{
"model": "gpt-image-2",
"prompt": "Design a high-resolution vertical character concept poster ... (此处粘贴任一条角色设计 Prompt)"
}'
实践建议：
- 替换参数：凡见
{argument name="..." default="..."}、[SUBJECT]、[TYPE ...]、「XXX」均为可替换位，批量场景只需替换占位符； - 比例控制：文档案例普遍直接把
16:9 / 9:16 / 4:5 / 4:5 vertical写进 Prompt，不依赖额外参数； - 多角色保真：需要多人同框时，优先参考 Case 19 的“数量规则 + 身份规则 + Do not 清单”结构；
- 成品一致性：需要连续产出同一角色时，可利用仓库列出的“跨代保持一致的角色生成”能力配合角色参考图，例如 Case 2、5、27 的设定卡流程；
- 对照验证：每个案例在 images 目录下都配有真实输出图（如 images/character_case20、images/character_case30），复现时可先看参考图校准预期。
十一、小结：26 个案例沉淀出的 8 条 Prompt 工程经验
- 分点列需求（Case 2/5）比长句更易被遵循，尤其适合“三视图 + 表情 + 装备 + 色板”这类多要素任务；
- 把画幅写进 Prompt（Case 5/7/8/19）是跨 API 保版式的稳妥做法；
- 参数占位符（Case 10/11/12/13/23）让一条 Prompt 具备模板复用价值；
- 分层组织长 Prompt（Case 18/22/24/27）——角色/场景/灯光/镜头各归其位，模型更容易逐段消化；
- “要什么 + 不要什么”双向约束（Case 21/25）比单向描述更能锁死风格；
- 多角色必须写数量与身份规则（Case 19），否则容易出现克隆或漏人；
- DoD / 验收标准（Case 9）适合对交付质量有硬性要求的场景；
- 中英日多语言混写可行（Case 2/3/5/15/22），GPT-Image-2 对说明性文本的多语种理解相当稳健。
如需继续深挖，可对照阅读仓库中的英文版 cases/character.md、中文版 cases/character_zh-CN.md 等 11 语种案例文件，以及 README_es.md 中的 API 快速上手章节，将上述模板直接接入自己的生成管线。
创作声明：本文部分内容由AI辅助生成（AIGC），仅供参考
