四格 AI 风景对比图

Goal: Create a wide comparison graphic showing how four prompt-writing AI model tiers produce the same landscape scene in GPT Image 2, with a dark editorial UI frame and Japanese headline text. Canvas: Extra-wide horizontal banner, approximately 2:1 aspect ratio, black/dark charcoal background, clean social-media comparison layout. Use thin black dividers between image columns and a compact top title area. Top text: Large white Japanese headline: {argument name="headline text" default="同じ構図・同じ画像AI（GPT Image 2）。プロンプトを“書いたAI”だけが違う4枚"}. Under it, add a smaller gray subtitle: {argument name="subtitle text" default="変えたのは「プロンプトを書くClaudeのモデル階層」だけ（左＝最も安い → 右＝最上位）"}. Layout: Four equal vertical landscape panels placed side by side, each with a small dark label bar at the top. Use exactly 4 panels, ordered left to right: 1. Label: “haiku ｜ 最も安い” — a dramatic semi-realistic anime landscape at golden sunset, high mountain range in the distance, layered misty valleys, wildflower-covered rocky foreground, a tiny solitary person standing on a ridge, warm orange-purple clouds, slightly over-embellished cinematic atmosphere. 2. Label: “sonnet” — cleaner and more elegant sunset mountain valley, broad sky with soft glowing cumulus clouds, golden light spilling across rolling hills, a small lone figure on a ridge path, balanced composition, polished anime background style. 3. Label: “opus” — high-quality but restrained sunset vista, deep blue sky fading to orange near the horizon, large illuminated cloud formations, grassy hillside foreground, distant lake or reflective water, tiny figure near the bottom center, premium but conventional scenic framing. 4. Label: “fable ｜ 最上位” — emotionally directed bright morning or late-afternoon landscape, turquoise-blue sky, towering luminous white clouds, sunbeams over a vast green valley, a small person in a white dress and hat standing in meadow grass at lower right, more narrative and poetic mood. Visual style: Refined Japanese anime background art blended with painterly cinematic lighting, highly detailed skies and atmospheric perspective. Keep all four panels similar in composition and subject matter so the comparison feels controlled, but vary mood and art direction per label. Use crisp white text, dark label bars, no logos, no watermark. Constraints: Preserve exactly 4 image panels and exactly these 4 visible panel labels. Make the title and subtitle legible. Keep the overall image looking like a posted comparison chart rather than a standalone landscape collage.

目标：创建一个宽幅对比图，展示四个提示词编写 AI 模型层级如何在 GPT Image 2 中生成相同的风景场景，并配有深色编辑 UI 框架和日语标题文本。 画布：超宽横向横幅，比例约为 2:1，黑色/深炭灰色背景，简洁的社交媒体对比布局。在图像列之间使用细黑色分隔线，并设置紧凑的顶部标题区域。 顶部文本：大号白色日语标题：{argument name="headline text" default="同じ構図・同じ画像AI（GPT Image 2）。プロンプトを“書いたAI”だけが違う4枚"}。下方添加较小的灰色副标题：{argument name="subtitle text" default="変えたのは「プロンプトを書くClaudeのモデル階層」だけ（左＝最も安い → 右＝最上位）"}。 布局：四个等宽的垂直风景面板并排排列，每个面板顶部都有一个深色小标签栏。使用 4 个面板，从左到右顺序如下： 1. 标签：“haiku ｜ 最も安い” — 充满戏剧性的半写实动漫风景，金色日落，远处高耸的山脉，层叠的雾气山谷，长满野花的岩石前景，山脊上站着一个微小的孤独身影，温暖的橙紫色云层，略显华丽的电影感氛围。 2. 标签：“sonnet” — 更简洁优雅的日落山谷，广阔的天空伴有柔和发光的积云，金色的光芒洒在连绵起伏的丘陵上，山脊小径上有一个孤独的小身影，构图平衡，精致的动漫背景风格。 3. 标签：“opus” — 高质量但内敛的日落远景，深蓝色天空向地平线渐变为橙色，巨大的发光云层，草地前景，远处有湖泊或反光水面，底部中心附近有一个微小的人物，高级但传统的风景构图。 4. 标签：“fable ｜ 最上位” — 情感导向的明亮清晨或傍晚风景，青蓝色天空，高耸的发光白云，阳光洒在广阔的绿色山谷上，右下角草地上站着一个身穿白裙戴帽子的微小人物，更具叙事性和诗意氛围。 视觉风格：精致的日本动漫背景艺术与绘画般的电影光影相结合，高度细致的天空和大气透视。保持四个面板的构图和主题相似，以确保对比的严谨性，但根据标签改变情绪和艺术方向。使用清晰的白色文本，深色标签栏，无 Logo，无水印。 约束：保留 4 个图像面板和这 4 个可见面板标签。确保标题和副标题清晰易读。使整体图像看起来像是一张发布的对比图表，而不是独立的风景拼贴画。