"action": "image_generation", "action_input": "一位美得令人惊艳的女性,留着精心编织的深色辫发,自信而优雅地坐着。她姿态从容,双腿优雅交叠,身穿一件精致的宝蓝色长袖中长连衣裙,剪裁合体,搭配不透明的黑色连裤袜和同色系的宝蓝色尖头平底鞋。她神情平静、安宁,散发出自信与力量。她位于一间奢华现代的顶层公寓客厅中央,室内陈设为高端极简风格,配有柔和的燕麦色墙面、流畅的中世纪现代家具,以及温润的抛光胡桃木点缀。灯光具有电影质感且十分专业,采用浅景深营造柔和的背景虚化效果。温暖的环境光捕捉到她辫发的纹理和连衣裙浓郁的色泽,柔和的主光则突出了她的面部轮廓。整体氛围精致、通透、充满高级时尚感,呈现出干净、考究的美学风格,以及清晰的高分辨率细节。"}



{ "direction": "A real SNS portrait with a 9:16 aspect ratio, the kind taken casually by a colleague with a smartphone during a break by the window of a modern office. Reference the soft transparency of the image, hair in backlight, a sligh

Goal: Systematically combine frontal, profile, three-quarter profile, and slight pitch angles, while maintaining facial proportions and identity stability. { "composition": { "layout": "5 columns x 3 rows, exactly 15 portrait panels"




2x2 grid, do this for 4 non copyrighted public space books, like Shahnameh CLASSIFY(input) → { domain: macro_diorama × conceptual_book_art × material_transmutation intent: awe + narrative_synthesis + tactile_hyperrealism (coll

Generate 6 independent vertical photos at once, each as a separate image, no collages, no grids, no storyboards. [Reference Image Division] Reference Image 1 (Original): The sole source of the background. Directly use the real background fr