IF 场景数量 >= 2:
    尝试调用 generate_4grid_location_images
    IF 场景数量 < 4:
        用全高黑色占位图片补齐到4个（提示词："A pure black background image, solid black color, 16:9 aspect ratio, full height"）
        顺序：先填充上方两个位置（Shot 1, Shot 2），再填充下方两个位置（Shot 3, Shot 4）
    ENDIF
    IF 4宫格生成失败:
        记录失败原因
        改用 generate_text_to_image 逐个生成
    ENDIF
ELSE IF 场景数量 == 1:
    调用 generate_text_to_image
ENDIF
 
IF 道具数量 >= 2:
    尝试调用 generate_4grid_prop_images
    IF 道具数量 < 4:
        用全高黑色占位图片补齐到4个（提示词："A pure black background image, solid black color, 16:9 aspect ratio, full height"）
        顺序：先填充上方两个位置（Shot 1, Shot 2），再填充下方两个位置（Shot 3, Shot 4）
    ENDIF
    IF 4宫格生成失败:
        记录失败原因
        改用 generate_text_to_image 逐个生成
    ENDIF
ELSE IF 道具数量 == 1:
    调用 generate_text_to_image
ENDIF

{
  "grid_layout": "2x2",
  "grid_aspect_ratio": "16:9",
  "global_watermark": "",
  "shots": [
    {
      "shot_number": "Shot 1",
      "prompt_text": "场景/道具1的完整提示词"
    },
    {
      "shot_number": "Shot 2",
      "prompt_text": "场景/道具2的完整提示词"
    },
    {
      "shot_number": "Shot 3",
      "prompt_text": "场景/道具3的完整提示词"
    },
    {
      "shot_number": "Shot 4",
      "prompt_text": "场景/道具4的完整提示词"
    }
  ]
}

┌─────────┬─────────┐
│ Shot 1  │ Shot 2  │  场景/道具1(左上) → 场景/道具2(右上)
│ (左上)  │ (右上)  │
├─────────┼─────────┤
│ Shot 3  │ Shot 4  │  场景/道具3(左下) → 场景/道具4(右下)
│ (左下)  │ (右下)  │
└─────────┴─────────┘

ABSOLUTELY NO text, NO watermark, NO letters, NO characters, NO words, NO signs, NO writing, NO typography, NO labels, NO captions, NO subtitles, NO Chinese characters, NO English text, NO numbers, NO logos, NO stamps, NO seals, completely text-free image, pure visual content without any written language

{
  "success": true,
  "task_id": 26,
  "name": "GPT Image 2 图片编辑",
  "computing_power": 2,
  "supported_sizes": ["1k", "2k", "4k"],
  "supported_ratios": ["1:1", "2:3", "3:2", "16:9", "9:16"],
  "supports_grid_image": true,
  "default_size": "1k",
  "max_size": "4k",
  "cost_per_image_default_size": 2,
  "cost_per_image_max_size": 2,
  "driver_hint": "提示词长度请控制在5000字符以内，超出将被API拒绝。请精简描述，避免冗余。"
}

1. 调用 read_world() 获取 visual_style
2. 检查 visual_style 中是否包含"写实"、"真实"、"照片"、"摄影"等关键词
3. 如果包含 → 绝对禁止在提示词中使用以下词汇：
   ❌ "anime"
   ❌ "reference sheet"
   ❌ "design reference"
   ❌ "illustration"
   
   必须使用以下词汇：
   ✅ "photography"
   ✅ "photograph"
   ✅ "architectural photography"
   ✅ "location shoot"

[画风描述], [时代环境], [色彩基调]. A professional architectural photography of [场景英文名] ([场景中文名]) from [作品/世界名称], shot with high-quality camera equipment. The location is captured without any people or objects present.
 
Wide-angle photograph showing the complete overview of [场景名] featuring [详细的环境描述，包括地形、建筑、植被、天气]. The scene displays [空间布局描述，包括主要区域、地标、道路]. Architectural style: [建筑风格详述]. Environmental features: [环境特征如光照、氛围、季节]. The location is completely empty with no people or objects. The overall mood conveys [基于场景特点的氛围描述]. Natural lighting and realistic textures are emphasized. No text, signs, or written language appear in the scene. ABSOLUTELY NO text, NO watermark, NO letters, NO characters, NO words, NO signs, NO writing, NO typography, NO labels, NO captions, NO subtitles, NO Chinese characters, NO English text, NO numbers, NO logos, NO stamps, NO seals, completely text-free image, pure visual content without any written language.

[画风描述], [时代环境], [色彩基调]. A professional location design reference sheet for [场景英文名] ([场景中文名]) from [作品/世界名称], set on a clean neutral background. The layout shows the location without any characters or props. Avoid including any text, words, or written characters in the image.
 
Large detailed illustration showing the complete overview of [场景名] featuring [详细的环境描述，包括地形、建筑、植被、天气]. The scene displays [空间布局描述，包括主要区域、地标、道路]. Architectural style: [建筑风格详述]. Environmental features: [环境特征如光照、氛围、季节]. The location is completely empty with no people or objects. The overall mood conveys [基于场景特点的氛围描述]. Do not include any text, signs, or written language in the scene. ABSOLUTELY NO text, NO watermark, NO letters, NO characters, NO words, NO signs, NO writing, NO typography, NO labels, NO captions, NO subtitles, NO Chinese characters, NO English text, NO numbers, NO logos, NO stamps, NO seals, completely text-free image, pure visual content without any written language.

[画风描述], [时代环境], [色彩基调]. A professional product photography collection of [道具英文名] ([道具中文名]) from [作品/世界名称], shot in studio lighting against a clean white backdrop. The prop is photographed alone without any hands, characters, or other objects.
 
Three-Angle Photography: Three high-quality photographs of [道具名] arranged horizontally. Front Photograph (Left): [道具正面照片的详细描述，包括形状、材质、颜色、纹理、装饰细节]. Side Photograph (Center): [道具侧面照片的详细描述，展现厚度、轮廓、连接部分]. Back Photograph (Right): [道具背面照片的详细描述，必须与正面保持一致的设计元素和比例]. 
 
Physical characteristics: [尺寸比例描述], [材质质感如金属光泽、木纹理、宝石透明度等], [表面装饰如雕刻、镶嵌、符文等], [磨损状态和使用痕迹]. All three photographs maintain perfect consistency in proportions, colors, materials, and decorative elements. The prop shows [基于功能和历史的整体状态描述]. Professional studio lighting highlights textures and details. No characters, hands, text, words, or other objects appear in any photograph. ABSOLUTELY NO text, NO watermark, NO letters, NO characters, NO words, NO signs, NO writing, NO typography, NO labels, NO captions, NO subtitles, NO Chinese characters, NO English text, NO numbers, NO logos, NO stamps, NO seals, completely text-free image, pure visual content without any written language.

[画风描述], [时代环境], [色彩基调]. A clean prop design reference sheet for [道具英文名] ([道具中文名]) from [作品/世界名称], set on a white background. The prop appears alone without any characters or hands holding it. Avoid including any text, words, or written characters in the image.
 
Three-View Display: Three precise technical views of [道具名] arranged horizontally. Front View (Left): [道具正面视图的详细描述，包括形状、材质、颜色、纹理、装饰细节]. Side View (Center): [道具侧面视图的详细描述，展现厚度、轮廓、连接部分]. Back View (Right): [道具背面视图的详细描述，必须与正面保持一致的设计元素和比例]. 
 
Physical characteristics: [尺寸比例描述], [材质质感如金属光泽、木纹理、宝石透明度等], [表面装饰如雕刻、镶嵌、符文等], [磨损状态和使用痕迹]. All three views must maintain perfect consistency in proportions, colors, materials, and decorative elements. The prop shows [基于功能和历史的整体状态描述]. No characters, hands, text, words, or other objects should appear in the image. ABSOLUTELY NO text, NO watermark, NO letters, NO characters, NO words, NO signs, NO writing, NO typography, NO labels, NO captions, NO subtitles, NO Chinese characters, NO English text, NO numbers, NO logos, NO stamps, NO seals, completely text-free image, pure visual content without any written language.

中文: "现代都市写实风格，高饱和度，强调奢侈品和豪宅的金属光泽与质感"
✅ 正确: "Photorealistic modern urban style, high saturation, emphasizing metallic luster and texture of luxury goods and mansions"
❌ 错误: "Semi-realistic anime style, modern urban setting, high contrast with golden metallic luster"
 
中文: "日系动漫风格，柔和色调"
✅ 正确: "Japanese anime style, soft color tones"
❌ 错误: "Photorealistic Japanese style"
 
中文: "古风写实，水墨意境"
✅ 正确: "Photorealistic ancient Chinese style with traditional ink painting aesthetics"
❌ 错误: "Ancient Chinese anime style"

[根据visual_style动态生成，如：Japanese anime style / Photorealistic style / Cartoon style], Ancient Chinese setting, Warm earthy color palette. A professional location design reference sheet for Moonlit Pavilion (月影亭) from The Scholar's Journey, set on a clean neutral background. The layout shows the location without any characters or props.
 
Large detailed illustration showing the complete overview of Moonlit Pavilion featuring a traditional Chinese octagonal pavilion built on a small artificial island in the center of a tranquil lotus pond, connected to the shore by an elegant arched stone bridge with carved dragon railings. The pavilion has upturned eaves with intricate bracket systems, red lacquered pillars with gold dragon motifs, and a dark green glazed tile roof topped with a golden finial. Surrounding the pond are weeping willows with drooping branches touching the water surface, carefully manicured rock gardens with ornamental stones, and winding cobblestone paths. The location is completely empty with no people or objects. The scene is illuminated by soft moonlight filtering through thin clouds, creating silver reflections on the water and casting gentle shadows. The overall mood conveys peaceful contemplation and scholarly refinement. Do not include any text, signs, or written language in the scene. ABSOLUTELY NO text, NO watermark, NO letters, NO characters, NO words, NO signs, NO writing, NO typography, NO labels, NO captions, NO subtitles, NO Chinese characters, NO English text, NO numbers, NO logos, NO stamps, NO seals, completely text-free image, pure visual content without any written language.

[根据visual_style动态生成，如：Japanese anime style / Photorealistic style / Cartoon style], Ancient Chinese setting, Warm earthy color palette. A clean prop design reference sheet for Scholar's Brush (文士之笔) from The Scholar's Journey, set on a white background. The prop appears alone without any characters or hands holding it.
 
Three-View Display: Three precise technical views of Scholar's Brush arranged horizontally. Front View (Left): The brush displayed vertically showing its full length, featuring an elegant bamboo handle carved with flowing cloud patterns in natural brown color with darker nodes and subtle grain texture, a silver metal ferrule engraved with ancient Chinese characters "知慧" (knowledge and wisdom), and a fine wolf hair brush tip with perfect point and rich black color. The carved cloud patterns spiral around the handle in flowing, organic curves that follow the natural bamboo segments. Side View (Center): The brush shown in profile revealing its cylindrical form and proportions, displaying the thickness of the bamboo handle (approximately 1.2cm diameter), the tapered ferrule connection, and the brush tip's gradual taper from thick base to fine point. The side view clearly shows the depth of the carved patterns and the ferrule's engraved characters. Back View (Right): The opposite side of the brush showing the continuation of the cloud pattern carving, maintaining perfect consistency with the front view in terms of bamboo color, node placement, and carving depth. The ferrule shows the same silver finish and the brush tip maintains identical shape and color.
 
Physical characteristics: approximately 25cm in total length with 15cm handle and 10cm brush head, lightweight yet balanced feel, smooth bamboo surface polished to a subtle sheen, intricate carving work showing masterful craftsmanship, slight wear marks on the grip area from years of scholarly use. All three views must maintain perfect consistency in proportions, colors, materials, and decorative elements. The brush shows signs of being a treasured tool of learning, well-maintained but bearing the gentle patina of age and frequent use. No characters, hands, or other objects should appear in the image. ABSOLUTELY NO text, NO watermark, NO letters, NO characters, NO words, NO signs, NO writing, NO typography, NO labels, NO captions, NO subtitles, NO Chinese characters, NO English text, NO numbers, NO logos, NO stamps, NO seals, completely text-free image, pure visual content without any written language.

步骤1: 统计需要生成的场景数量 = N_locations, 道具数量 = N_props
步骤2: 
  IF N_locations >= 2:
    # 优先尝试4宫格批量生成
    调用函数 = "generate_4grid_location_images"
    IF N_locations < 4:
      补齐数量 = 4 - N_locations
      location_names = [实际场景列表] + ["placeholder"] * 补齐数量
      prompts = [实际提示词列表] + ["A pure black background image, solid black color, 16:9 aspect ratio, full height"] * 补齐数量
      # 顺序：Shot 1, Shot 2, Shot 3, Shot 4（先上排后下排）
    ELSE:
      location_names = [前4个场景]
      prompts = [对应提示词]
    ENDIF
    参数 = {location_names: location_names, prompts: prompts}
    
    # 如果4宫格生成失败，改用逐个生成
    IF 4宫格生成失败:
      记录失败原因
      FOR 每个场景:
        调用 generate_text_to_image() 单独生成
      ENDFOR
    ENDIF
  ELSE IF N_locations == 1:
    调用函数 = "generate_text_to_image"
    处理单个场景
  ENDIF
  
  IF N_props >= 2:
    # 优先尝试4宫格批量生成
    调用函数 = "generate_4grid_prop_images"
    IF N_props < 4:
      补齐数量 = 4 - N_props
      prop_names = [实际道具列表] + ["placeholder"] * 补齐数量
      prompts = [实际提示词列表] + ["A pure black background image, solid black color, 16:9 aspect ratio, full height"] * 补齐数量
      # 顺序：Shot 1, Shot 2, Shot 3, Shot 4（先上排后下排）
    ELSE:
      prop_names = [前4个道具]
      prompts = [对应提示词]
    ENDIF
    参数 = {prop_names: prop_names, prompts: prompts}
    
    # 如果4宫格生成失败，改用逐个生成
    IF 4宫格生成失败:
      记录失败原因
      FOR 每个道具:
        调用 generate_text_to_image() 单独生成
      ENDFOR
    ENDIF
  ELSE IF N_props == 1:
    调用函数 = "generate_text_to_image"
    处理单个道具
  ENDIF
步骤3: 执行调用

# 错误！未尝试4宫格就直接逐个生成：
for location in ["场景1", "场景2"]:
    generate_text_to_image(item_name=location, item_type=2, ...)  # ❌ 错误！应该先尝试4宫格
# 错误！未尝试4宫格就直接逐个生成：
for prop in ["道具1", "道具2"]:
    generate_text_to_image(item_name=prop, item_type=3, ...)  # ❌ 错误！应该先尝试4宫格

# 正确！当有4个场景时，优先使用4宫格：
result = generate_4grid_location_images(
    location_names=["场景1", "场景2", "场景3", "场景4"],
    prompts=[提示词1, 提示词2, 提示词3, 提示词4]
)
# 如果4宫格失败，降级为逐个生成：
if not result.get("success"):
    for location, prompt in zip(locations, prompts):
        generate_text_to_image(item_name=location, item_type=2, prompt=prompt)
# ✅ 正确！
# 正确！当有2个场景时（用占位符补齐）：
result = generate_4grid_location_images(
    location_names=["场景1", "场景2", "placeholder", "placeholder"],
    prompts=[提示词1, 提示词2, "pure black background", "pure black background"]
)
if not result.get("success"):
    # 降级为逐个生成
    for location, prompt in [("场景1", 提示词1), ("场景2", 提示词2)]:
        generate_text_to_image(item_name=location, item_type=2, prompt=prompt)
# ✅ 正确！
# 正确！当有3个道具时（用占位符补齐）：
result = generate_4grid_prop_images(
    prop_names=["道具1", "道具2", "道具3", "placeholder"],
    prompts=[提示词1, 提示词2, 提示词3, "pure black background"]
)
if not result.get("success"):
    # 降级为逐个生成
    for prop, prompt in [("道具1", 提示词1), ("道具2", 提示词2), ("道具3", 提示词3)]:
        generate_text_to_image(item_name=prop, item_type=3, prompt=prompt)
# ✅ 正确！