jeffstric/zjt/character-image-designer

对于每个需要处理的角色：
  1. 检查 reference_image 是否为空
  2. 检查 default_voice 是否为空
  3. IF reference_image 为空:
       生成参考图像
  4. IF default_voice 为空:
       生成参考音色  # 必须执行，不可跳过
  5. 在报告中同时报告图像和音色的生成状态

IF 角色数量 >= 2:
    尝试调用 generate_4grid_character_images
    IF 角色数量 < 4:
        用黑色占位图片补齐到4个
    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 角色数量 == 1:
    调用 generate_text_to_image
ENDIF

{
  "grid_layout": "2x2",
  "grid_aspect_ratio": "16:9",
  "global_watermark": "",
  "shots": [
    {
      "shot_number": "",
      "prompt_text": "角色1的面部特写+三视角提示词"
    },
    {
      "shot_number": "",
      "prompt_text": "角色2的面部特写+三视角提示词"
    },
    {
      "shot_number": "",
      "prompt_text": "角色3的面部特写+三视角提示词"
    },
    {
      "shot_number": "",
      "prompt_text": "角色4的面部特写+三视角提示词"
    }
  ]
}

┌─────────┬─────────┐
│ Shot 1  │ Shot 2  │  角色1(左上) → 角色2(右上)
│ (左上)  │ (右上)  │
├─────────┼─────────┤
│ Shot 3  │ Shot 4  │  角色3(左下) → 角色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

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"
   ❌ "grid paper"
   ❌ "borders"
   ❌ "illustrations"
   ❌ "sketches"
   
   必须使用以下词汇：
   ✅ "photography portfolio"
   ✅ "photographs"
   ✅ "studio lighting"
   ✅ "shots"
   ✅ "portraits"

world_info = read_world()
visual_style = world_info.get('visual_style', '')
# 明确输出判断结果
print(f"检测到的画风: {visual_style}")
if any(keyword in visual_style for keyword in ['写实', '真实', '照片', '摄影']):
    template_type = "REALISTIC_PHOTOGRAPHY"
    print("✅ 使用写实风格模板")
elif any(keyword in visual_style for keyword in ['动漫', '二次元', '卡通']):
    template_type = "ANIME_REFERENCE"
    print("✅ 使用动漫风格模板")
else:
    template_type = "REALISTIC_PHOTOGRAPHY"
    print("⚠️ 未明确指定，默认使用写实风格模板")

# 生成提示词后，必须进行以下验证
if template_type == "REALISTIC_PHOTOGRAPHY":
    # 写实风格提示词必须包含的关键词
    required_keywords = ["photography portfolio", "close-up facial portrait", "front angle", "side profile", "back view", "cropped at the neck, no head"]
    forbidden_keywords = ["anime", "reference sheet", "grid paper", "borders"]
    
    # 检查是否包含必需关键词
    if not all(keyword in prompt for keyword in ["close-up facial portrait", "front angle", "side profile", "back view", "cropped at the neck, no head"]):
        raise Error("❌ 写实风格提示词缺少必需的面部特写/三视角/三视图无头部描述！")
    
    # 检查是否包含禁止关键词
    if any(keyword in prompt for keyword in forbidden_keywords):
        raise Error("❌ 写实风格提示词中不能包含动漫术语！")
    
    print("✅ 写实风格提示词验证通过")
elif template_type == "ANIME_REFERENCE":
    # 动漫风格提示词必须包含的关键词
    required_keywords = ["reference sheet", "close-up facial portrait", "front view", "side profile", "back view", "cropped at the neck, no head"]
    
    if not all(keyword in prompt for keyword in ["close-up facial portrait", "front view", "side profile", "back view", "cropped at the neck, no head"]):
        raise Error("❌ 动漫风格提示词缺少必需的面部特写/三视角/三视图无头部描述！")
    
    print("✅ 动漫风格提示词验证通过")

# 伪代码示例
world_info = read_world()
visual_style = world_info.get('visual_style', '')
# 判断画风类型
if any(keyword in visual_style for keyword in ['写实', '真实', '照片', '摄影']):
    # 使用写实风格模板
    template_type = "realistic_photography"
    # 提示词中使用：photography portfolio, studio lighting, photographs, shots 等摄影术语
    # 绝对禁止使用：reference sheet, grid paper, borders 等动漫设定图术语
elif any(keyword in visual_style for keyword in ['动漫', '二次元', '卡通']):
    # 使用动漫风格模板
    template_type = "anime_reference"
    # 提示词中使用：reference sheet, grid paper, borders 等动漫设定图术语
else:
    # 默认使用写实风格
    template_type = "realistic_photography"

[画风描述], [时代环境], [色彩基调]. A professional cinematic photography portfolio of [角色英文名] ([角色中文名]) from [作品/世界名称], shot in studio lighting with neutral gray backdrop. On the far left, a high-resolution close-up facial portrait of [角色名] (a head-and-shoulders shot highlighting detailed facial features, skin texture, eye color, hair framing the face and a clear facial expression), occupying approximately the left quarter of the image. To its right, three full-body photographs arranged in a horizontal row from left to right: [角色名] from front angle (body facing the camera, cropped at the neck, no head), side profile (90-degree turn to the left, cropped at the neck, no head), and back view (facing away from camera, showing only the back of the head, no facial features visible). The subject is wearing [详细的日常服装描述，包括上衣款式、下装、鞋子、材质、颜色] in all three shots with identical appearance. Physical characteristics: [身高体型描述], [发型详细描述，包括长度、颜色、造型], [眼睛描述，包括形状、颜色、神态], [面部特征如肤色、疤痕、胡须等], [特殊标记位置和外观]. Facial expression: [基于性格的默认表情，仅体现在左侧特写中]. CRITICAL LAYOUT RULE: the close-up facial portrait on the far left is the ONLY view showing the character's face; the front and side full-body views to its right must be strictly HEADLESS (cropped at the neck, no head, no face, no hair rendered at all), while the back view shows only the back of the head with no facial features. 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.