OpenZhenZhen AI 提示词画廊 · gpt-image-2.5 · image

原图主体与风格派上下等分重构插画

De Stijl Split Reconstruction Illustration of the Original Subject

竖版上下双拼插画：上半部忠实提炼上传原图的单一主体，下半部将同一主体转译为红黄蓝、黑线与白色块面的De Stijl风格；两部分等分，背景简化且主体完整。

原图主体与风格派上下等分重构插画 — AI image creation
原图主体与风格派上下等分重构插画

提示词 / Prompt

中文提示词

创作一张经过精心设计的竖版上下分屏插画，而非简单拼贴。画布严格分为上下两个等面积部分，比例为1:1。必须使用用户上传的原图作为唯一主体来源。上半部：从原图中识别一个明确的核心主体，例如主要人物、猫、狗或其他常见宠物；重点提炼主体，不让建筑或复杂环境喧宾夺主。保留主体最重要的识别信息、构图与可辨认姿态。若主体是人物，只呈现一位明确成年、20岁以上的人物，使用固定身份ref-10a-v1，不新增其他人物；若主体是动物，则准确保留原动物的种类、毛色与花纹、体型和五官，不套用人类身份。适度简化背景干扰，主体完整清晰。下半部：将同一主体重构为可辨识的De Stijl风格派视觉语言，采用有序的横竖黑线、红黄蓝原色块与白色区域；结构清晰、形状扁平，层级明确并留有充足负空间。将主体轮廓、姿态、服饰或毛发花纹关系及周围负空间理性纳入几何系统。下半部须一眼可辨与上半部为同一原图主体，并带有手作印刷/绘画质感。全图保持单一协调画风，避免混入无关风格、过度平滑、无意义细节堆砌或随机装饰。除非主题明确需要，不添加文字、标志、水印、边框或UI；不要制作多图拼贴。

English prompt

Create a carefully designed portrait-oriented split-screen illustration, not a simple collage. Divide the canvas into equal upper and lower panels (a strict 1:1 split). The uploaded original image is required and is the sole source for the main subject. Upper panel: identify one clear focal subject from the image, such as its principal person, cat, dog or other common pet; prioritize the subject over architecture or a complex setting. Preserve its most identifying information, composition and recognizable pose. If the focal subject is human, depict one clearly adult person, age 20+, using fixed identity ref-10a-v1; do not introduce other people. If it is an animal, retain that exact animal's species, fur color/pattern, body shape and facial features, and do not apply a human identity. Refine the background by simplifying distractions while keeping the subject complete and clear. Lower panel: reconstruct that same subject in recognizable De Stijl visual language: orderly horizontal and vertical black lines, red/yellow/blue primary-color blocks and white areas; clear geometric structure, flat shapes, controlled hierarchy and generous negative space. Rationally translate the subject's silhouette, pose, clothing or fur-pattern relationships, and surrounding negative space into the geometric system. Keep the lower panel instantly recognizable as the same source subject, with a handmade print/paint feel. Use one coherent style only; avoid unrelated styles, excessive smoothing, meaningless detail and random decoration. Unless the theme specifically requires them, do not add text, logos, watermark, border or UI; do not create a multi-image collage.

查看 API 模型

探索更多图片和视频提示词 · 生成图片 · 生成视频 · 在 AI 画布中创作