生成一张《纸上成真 · 神话绮装》风格图片，4:5 竖版。 — GPT Image 2 Prompt

1 / 4

生成一张《纸上成真 · 神话绮装》风格图片，4:5 竖版。画面是高级东方神话高定服饰设定稿：米白色手稿纸夹在木质画板上，放在温暖木色设计工作台。桌面有布料小样… — AI-generated image by @MrLarus using GPT Image 2

Larus Canus

Prompt

生成一张《纸上成真 · 神话绮装》风格图片，4:5 竖版。 画面是高级东方神话高定服饰设定稿：米白色手稿纸夹在木质画板上，放在温暖木色设计工作台。桌面有布料小样、配色色卡、铅笔、针管笔、陶瓷茶杯、玉饰、金属配件和小型服饰结构草图。 主体是一位成年中国神话女性人物【女娲 / 洛神 / 嫦娥 / 九天玄女】，从纸面服饰设定稿中苏醒。人物非常写实，像真实人物摄影，皮肤白皙自然，有真实皮肤纹理，不要动漫感、硅胶感、塑料皮肤。 人物眼神要和画面有互动，可以看向镜头、看向创作者，或看向右下角正在绘制的铅笔区域。上半身必须明显从纸面立起来，有真实体积、空间高度和自然投影，一只手轻撑纸面，形成"纸上成真"的错觉。 人物比例修长但自然，约 7–7.5 头身，上半身清楚可见，胸腔、腰部、胯部和腿部比例协调。腿部偏写实呈现，修长自然，不要小腿过粗，不要下半身过短或过长。 服饰为东方神话高定风格，主色调为【玉白青绿淡金 / 月白浅蓝银灰 / 霜银浅雾蓝 / 玄黑暗金深红】。服装要有连续统一的颜色层次，不要斑驳杂乱。 从上到下呈现：写实人物与服装主体 → 腿部和中段裙装偏写实 → 裙角、披帛尾端、飘带末端逐渐变成淡彩手绘 → 右下角铅笔正在画的区域保留未完成铅笔素描。 纸面加入少量手稿标注：LOOK 01、silhouette、ribbon flow、skirt movement、ornament detail，也可加入中文：裙摆走势、发饰位置、披帛结构、金饰位置。 真实摄影感，高级东方审美，神话人物高定设定稿，纸张纹理清晰，阴影真实，构图干净。避免卡通、动漫脸、硅胶皮肤、文字乱码、人体比例错误、上半身太小、腿部畸形、全身过度手绘、颜色斑驳、低清晰度。

65.6K views · 725 characters

How to use this prompt

Copy the full prompt with the Copy Prompt button.
Open GPT Image 2 in ChatGPT and paste it — attach your own photo as a reference to apply the style to it.
Generate, then tweak one detail at a time (subject, colors, framing) to remix the result.

FAQ

Where do I use this GPT Image 2 prompt?: Open ChatGPT (image generation is available on free and paid plans), paste the prompt, and send. GPT Image 2 follows detailed, multi-part instructions well, so you can be specific about layout and composition.
Can I use my own photo with GPT Image 2?: Yes. Attach your photo in ChatGPT first, then paste the prompt — GPT Image 2 edits from a reference image, so it restyles your photo while keeping the subject. Keep the prompt focused on the change you want.
Is GPT Image 2 free to use?: Image generation is available to free ChatGPT users with limits; paid plans raise those limits. Copying and reusing this prompt is free either way.

More like this

@Chaemate_

@Kashberg_0

@LinusEkenstam

@johnAGI168

View on X

Prompt details from the source text

These notes expose only instructions that are observable in the shared prompt or source metadata. A field is marked as not specified when the source does not state it; no output has been inferred or rewritten.

Use and setup

Use case: Recorded category: Portraits.
Reference image: No reference-image requirement is stated.
Output ratio or size: Explicit size/aspect signal: 4:5.
Variables or slots: No reusable variable or slot is stated.

Prompt breakdown

Subject: Not specified in the source prompt.
Composition: Not specified in the source prompt.
Camera: Not specified in the source prompt.
Lighting: Not specified in the source prompt.
Material: Not specified in the source prompt.
Text or typography: Not specified in the source prompt.
Negative requirements: Not specified in the source prompt.

Recorded model:GPT Image 2(source metadata; verify fit in your own workflow)