几何伴游说明书风上下对比海报

Convert any real photo into a "Geometric Companion Manual style" up-down comparison transformation poster. 
Composition/framing: Strict up-down partitioning, with the upper part of the image being about 45% high-fidelity real photography of the input photo, and the lower part about 55% being a geometric manual re-drawing of the same subject; the horizontal boundary is clear but natural, side-by-side, nine-grid, and multi-image collages are prohibited. The lower part changes the scale and information density of the subject, do not mechanically trace the original composition of the photo. 
Style/medium: Paper instruction cards, travel observation pages, restrained geometric collage and hand-printed diagrams. Deconstruct the main subject in the photo into a small number of flat geometric planes, black outlines of varying thickness, and limited 2–4 color blocks; retain the identity, quantity, action direction, and most critical structure of the subject. Use arrows, dotted lines, numbered dots, partial enlargement boxes, short labels, and concise structural lines to explain visible features, forming an instructional relationship like "how to observe/how to identify." Annotations must point to actual visible parts of the subject or spatial relationships, do not add objects that do not exist in the photo. 
Scene/backdrop: Use a warm off-white paper background for the lower half, with slight paper fibers, uneven printing, and low-contrast grain; colors are extracted and compressed from the input photo, do not fix templates for blue seas or dogs. Leave breathing space between the subject and the instruction boxes. 
Text: Include only 1 short English title, 1 material/observation field, and 1 extremely short instruction; text is like labels printed on paper cards, avoid long paragraphs, fake Chinese, URLs, QR codes, brands, and watermarks. If the photo subject is different, the title and fields must be replaced according to the facts of the photo. 
Constraints: Maintain the identity, quantity, posture, action direction, key structure, lighting direction, and main spatial relationships of the subject in the real photo; the upper part cannot be drawn as an illustration, the subject cannot be replaced, and the photo cannot be blurred or reconstructed. The lower part must still be recognizable as the same subject from the upper part at a glance, and clearly reflect geometric decomposition, arrow annotations, and the manual page mechanism. Do not show "original image/AI image" badges. 
Avoid: Only color adjustment, only adding grain, ordinary cartoon filters, realistic photos filling the lower area, meaningless decorative stickers, fixed dog/beach elements, overly complex engineering drawings, dense and hard-to-read small text, extra figures, and irrelevant landmarks.

将任意真实照片转化为“几何伴游说明书风”的上下对比转绘海报。
Composition/framing: 严格上下分区，画面上半部约 45% 为输入照片的高保真真实摄影，下半部约 55% 为同一主体的几何说明书转绘；水平分界清楚但自然，禁止左右并排、九宫格和多图拼贴。下半部改变主体尺度与信息密度，不要把照片原构图机械描一遍。
Style/medium: 纸张说明卡、旅行观察页、克制的几何拼贴与手工印刷图解。把照片中的主体拆成少量平面几何面、粗细不一的黑色轮廓线和有限的 2–4 色块；保留主体身份、数量、动作方向和最关键的结构。用箭头、虚线、编号圆点、局部放大框、短标签和简洁结构线解释可见特征，形成像“如何观察/如何识别”一样的说明关系。标注必须指向真实可见的主体部件或空间关系，不添加照片中不存在的对象。
Scene/backdrop: 下半部使用温暖的米白纸张底，轻微纸纤维、印刷不齐和低对比颗粒；颜色从输入照片提取并压缩，不固定蓝海或狗的模板。主体与说明框之间留出呼吸空间。
Text: 只加入 1 个简短英文标题、1 个材料/观察字段和 1 句极短说明；文字像印在纸卡上的标签，避免长段落、假中文、网址、二维码、品牌和水印。若照片主题不同，标题和字段必须根据照片事实替换。
Constraints: 保持真实照片中的主体身份、数量、姿态、动作方向、关键结构、光照方向和主要空间关系；上半部不能画成插画、不能替换主体、不能模糊或重构照片。下半部必须仍能一眼对应上半部同一主体，并明确体现几何分解、箭头标注和说明页机制。不要出现“原图/AI图”角标。
Avoid: 仅调色、仅加颗粒、普通卡通滤镜、写实照片占满下半区、无意义装饰贴纸、固定狗/海滩元素、过度复杂的工程制图、密集难读的小字、额外人物和无关地标。