返回创作内容库

三组角色资产对比项目

生成一个 16:9 的对比项目,展示两列模型下三组写实角色资产表,适用于视觉 AI 输出对比。

gpt-image-2 · 人像 / 自拍 · 摄影 · 极简主义 · 游戏素材 · 角色

试一试复制喜欢的提示词,在 AnchorDot 中开始创作。
三组角色资产对比项目三组角色资产对比项目

中文提示词

目标:创建一个深色背景的对比项目,展示三位职场装扮人物的角色资产表,并排对比两个图像生成系统。 画布:16:9 横向演示图像,分辨率约为 2048×1152,采用深海军蓝黑色背景,简洁的白色字体,以及以 3 行 2 列网格排列的 6 个大型白色图像面板。 标题文本:在左上角放置大号中文标题 {argument name="headline text" default="3组 女生人物资产 · 同提示词对比"}。在其下方添加一行较小的副标题:{argument name="comparison subtitle" default="左:历史 GPT Image 2 右:ChatGPT Images 2.5 网页"}。 布局:使用 6 个带有细间距的矩形白色面板:左列代表旧模型,右列代表新模型。共有 3 行角色:Clara、Julian 和 Evelyn。每一行在左侧面板上方都有一个小的中文行标签:“纯文字角色资产 - Clara”、“纯文字角色资产 - Julian”和“纯文字角色资产 - Evelyn”。6 个面板中的每一个都在白色摄影棚背景下包含 4 个角色描绘:左侧为 1 个大型头部及肩部特写,右侧为 3 个较小的全身转向视图:正面视图、侧面视图和背面视图。可见角色描绘总数:正好 24 个。 第 1 行主体细节:Clara,一位年轻白皙女性,留着长波浪深棕色头发,淡雅自然妆容,表情中性严肃,身穿奶油色罗纹高领毛衣、棕色短款西装外套、高腰阔腿奶油色长裤、黑色腰带,搭配低跟鞋。在每个 Clara 面板中展示 4 个视图:头部特写、全身正面、全身侧面、全身背面。 第 2 行主体细节:Julian,一位年轻苗条白皙男性,留着浅棕色/金色侧分发型,佩戴矩形眼镜,表情冷静严肃,身穿米白色亚麻西装、白色开领衬衫、修身长裤、棕色腰带,搭配棕褐色乐福鞋。在每个 Julian 面板中展示 4 个视图:头部特写、全身正面、全身侧面、全身背面。 第 3 行主体细节:Evelyn,一位中年女性,中等肤色,深色头发梳成低发髻,表情沉稳中性,身穿黑色商务西装、深灰色衬衫,佩戴小金项链、小耳环,黑色长裤,搭配黑色平底鞋。在每个 Evelyn 面板中展示 4 个视图:头部特写、全身正面、全身侧面、全身背面。 视觉风格:写实摄影风格的角色资产表,干净的摄影棚灯光,每个面板内均为白色无缝背景,一致的专业服装造型,中性表情,除深色项目背景和标签外,无装饰性图形。左右两列应看起来相似但不完全相同,如同在对比两个系统的输出。 页脚文本:在左下角添加两条中文注释:{argument name="footer note one" default="历史成品 2048×1152 / 网页新图 1672×941 · 仅排版缩放,无修图、无超分。"} 和 {argument name="footer note two" default="提示词正文相同,两页端仅补充16:9比例。"}。 约束:保持 3 行 2 列的精确对比布局,正好 3 个命名角色,正好 6 个面板,每个面板正好 4 个描绘,无水印,无额外徽标,且无其他角色。

原始提示词

Goal: Create a dark-background comparison board showing character asset sheets for three office-dressed people, comparing two image-generation systems side by side. Canvas: 16:9 landscape presentation image, approximately 2048×1152, with a deep navy-black background, clean white typography, and six large white image panels arranged in a 3-row by 2-column grid. Header text: At the top left, place the large Chinese title {argument name="headline text" default="3组 女生人物资产 · 同提示词对比"}. Under it, add a smaller subtitle line: {argument name="comparison subtitle" default="左:历史 GPT Image 2 右:ChatGPT Images 2.5 网页"}. Layout: Use exactly 6 rectangular white panels with thin spacing: left column represents the older model, right column represents the newer model. There are exactly 3 character rows: Clara, Julian, and Evelyn. Each row has a small Chinese row label above the left panel: “纯文字角色资产 - Clara”, “纯文字角色资产 - Julian”, and “纯文字角色资产 - Evelyn”. Each of the 6 panels contains exactly 4 character depictions on a white studio background: 1 large cropped head-and-shoulders portrait on the left, plus 3 smaller full-body turnaround views on the right: front view, side profile view, and back view. Total visible character depictions: exactly 24. Row 1 subject details: Clara, a young fair-skinned woman with long wavy dark brown hair, soft natural makeup, neutral serious expression, wearing a cream ribbed turtleneck, brown cropped blazer, high-waisted wide-leg cream trousers, a black belt, and low heels. In each Clara panel show 4 views: close portrait, full-body front, full-body side, full-body back. Row 2 subject details: Julian, a young slim fair-skinned man with light brown/blond swept hair, rectangular glasses, calm serious expression, wearing an off-white linen suit, white open-collar shirt, narrow trousers, brown belt, and tan loafers. In each Julian panel show 4 views: close portrait, full-body front, full-body side, full-body back. Row 3 subject details: Evelyn, a middle-aged woman with medium tan skin, dark hair pulled back into a low bun, composed neutral expression, wearing a black business suit, dark gray blouse, small gold necklace, small earrings, black trousers, and black flats. In each Evelyn panel show 4 views: close portrait, full-body front, full-body side, full-body back. Visual style: Realistic photography-like character asset sheets, clean studio lighting, white seamless backgrounds inside each panel, consistent professional wardrobe styling, neutral expressions, no decorative graphics except the dark board and labels. The left and right columns should be visibly similar but not identical, as if comparing outputs from two systems. Footer text: At the bottom left, add two small Chinese notes: {argument name="footer note one" default="历史成品 2048×1152 / 网页新图 1672×941 · 仅排版缩放,无修图、无超分。"} and {argument name="footer note two" default="提示词正文相同,两页端仅补充16:9比例。"}. Constraints: Keep the exact 3-row by 2-column comparison layout, exactly 3 named characters, exactly 6 panels, exactly 4 depictions per panel, with no watermark, no extra logos, and no additional characters.