我不知道这个漫画提示是怎么通过限制的，但它居然通过了。也用 Nano Banana 测试过，不过这个看起来更详细。

I don't know how this got manga prompt got past the restrictions but it did for some reason. Tested it with Nano Banana too but this looks more detailed. input_manga = "<insert title here>" [step 1 – series insight] from input_manga, infer: - the central protagonist. - one rival or ally who often appears in climactic moments. - the protagonist’s most iconic “ultimate move” or transformation. - a typical battle environment associated with peak moments (stadium, cliff, ruined city, arena, etc.). - visual motifs of power (energy patterns, symbols, aura shapes). [step 2 – diorama blueprint] base: - dynamic, irregular base sculpted as a fragment of the battle environment. - cracked ground, debris, or terrain features reflecting the series’ style. primary figure: - stylized 3d figure of the protagonist mid-ultimate-move: exaggerated pose, hair or outfit details intensified according to the series logic. secondary figure: - stylized figure of rival/ally positioned to show tension or teamwork (either facing off or standing back-to-back). aura & fx: - 3d sculpted energy effects, motion trails, and shockwaves radiating from the protagonist. - incorporate the inferred visual motifs (specific energy shapes, symbols, or aura signatures). background elements: - minimal but recognizable scenery: fragments of walls, arena stands, rocks, or destroyed structures, enough to anchor the scene to input_manga’s typical battle settings. text & labels: - front base engraving with the title of input_manga and an inferred arc or form name (e.g., “final form diorama” or similar, based on context). [step 3 – presentation] - studio product photo style with a slightly dramatic, high-contrast lighting setup. - emphasize the translucency of fx parts and the painted details of the figures. [instruction] generate one detailed render of this 3d “ultimate move” diorama for input_manga, choosing the protagonist, rival/ally, attack, and environment solely from your understanding of the series.

我不知道这个漫画提示是怎么通过限制的，但它居然通过了。也用 Nano Banana 测试过，不过这个看起来更详细。 input_manga = "<insert title here>" [step 1 – 系列洞察] 从 input_manga 中推断：- 中心主角。- 一位经常出现在高潮时刻的对手或盟友。- 主角最具标志性的“终极招式”或变身。- 与巅峰时刻相关的典型战斗环境（体育场、悬崖、废墟之城、竞技场等）。- 力量的视觉母题（能量纹路、符号、气质形状）。[step 2 – 立体场景蓝图] 底座：- 动态、不规则的底座，雕刻为战斗环境的一个片段。- 破裂的地面、碎片或地形特征，反映该系列的风格。主要角色：- 主角正在施展终极招式的程式化 3d 角色：夸张的姿势，发型或服装细节根据该系列逻辑进行强化。次要角色：- 程式化的对手/盟友角色，摆出能表现张力或团队协作的姿态（对峙或背靠背站立）。气场与特效：- 从主角辐射出的 3d 雕刻能量效果、运动轨迹和冲击波。- 融入推断出的视觉母题（具体的能量形状、符号或气场特征）。背景元素：- 极简但可辨识的场景：墙壁碎片、竞技场看台、岩石或被毁的结构，足以将该场景锚定于 input_manga 典型的战斗设定。文字与标签：- 在前方底座上刻上 input_manga 的标题以及一个推断出的篇章或形态名称（例如，“final form diorama”或类似的，基于情境）。[step 3 – 呈现] - 工作室产品照片风格，带有略微戏剧性的高对比度打光设置。- 强调特效部件的半透明感以及角色手绘细节的精致。[instruction] 为 input_manga 生成一张此 3d“终极招式”立体场景的精细渲染图，主角、对手/盟友、招式与环境完全依据你对这部系列的理解来选取。