feat: initial release v0.3.0

This commit is contained in:
saturn
2026-03-08 03:15:27 +08:00
commit 881ed44996
1311 changed files with 225407 additions and 0 deletions

View File

@@ -0,0 +1,31 @@
# Character Image To Description Prompt
Analyze the provided character image and write one detailed English visual description for image generation.
## Required content
Include all of the following:
1. Gender and approximate age range
2. Hair style and hair color
3. Face shape and facial features
4. Body build and silhouette
5. Clothing style and clothing details
6. Accessories and signature details
7. Overall style keywords
## Missing-part completion
If the image only shows part of the body, infer the missing parts consistently.
- Missing lower body: infer matching pants/skirt style
- Missing shoes: infer shoes that fit the outfit
- Missing accessory details: infer a few reasonable accessories
## Forbidden content
Do not mention:
- Skin color
- Eye color
- Facial expression
- Pose or action
- Background
## Output format
Return one plain English paragraph only (about 120-220 words).
Do not return markdown, bullets, titles, or JSON.

View File

@@ -0,0 +1,36 @@
# 图片反推角色描述提示词
请分析这张角色图片,生成一段详细的角色外貌描述(用于 AI 图片生成)。
## 输出要求
生成一段完整的角色视觉描述,包含以下要素:
1. 性别和年龄段(如:约二十五岁的男性)
2. 发型发色(如:黑色短发、微卷的棕色长发)
3. 脸型五官特征(如:剑眉星目、高鼻梁、薄唇)
4. 体态身材(如:身形修长、体格健壮)
5. 服装风格(如:深蓝色西装、白色衬衫、皮质腰带)
6. 配饰特征(如:左手戴银色手表、胸前别金色胸针)
7. 整体气质关键词(如:精英气质、禁欲系、高冷、温柔暖男)
## 缺失内容补齐规则
如果参考图只展示了部分身体(如上半身、头像),请根据已有信息合理推断并补全:
- **缺少下半身**:根据上衣风格推断裤装/裙装类型(如西装上衣配深蓝色西裤、休闲上衣配牛仔裤)
- **缺少鞋子**:根据整体穿搭风格推断鞋款(如正装配皮鞋、休闲装配运动鞋或帆布鞋)
- **缺少配饰细节**:根据角色气质合理添加配饰(如商务风配手表、休闲风配手环)
## 禁止描写
- 皮肤颜色
- 眼睛颜色
- 表情
- 动作
- 背景
- 姿势
## 输出格式
一段连贯的描述文字约200-300字直接可用于图片生成提示词。
只返回描述文字,不要有任何标题、序号或其他格式。

View File

@@ -0,0 +1,22 @@
# Reference Image To Character Sheet Prompt (img2img)
Use the provided reference image to extract face traits, hairstyle, body shape, and outfit structure.
## Style priority
Follow the user-selected style instruction as the highest-priority rule.
Use reference images only to preserve character identity traits (face, hairstyle, body shape, outfit structure), and do not let reference style override the requested style.
Only if no explicit style instruction is provided, you may preserve the original visual style.
## Generation rules
1. Ignore the original image color cast and lighting defects
2. Use clean, soft studio lighting
3. Keep natural, aesthetically correct body proportions
4. Do not copy blur, noise, compression artifacts, or defects
5. Output must be clear, sharp, and production-quality
6. Character expression should be neutral and calm, looking at camera
## Missing-part completion
If only half body or partial body is visible, infer and complete hidden parts consistently:
- Infer matching lower-body clothing
- Infer suitable footwear
- Infer missing hands/arms in a natural way

View File

@@ -0,0 +1,26 @@
# 参考图转角色设定图提示词(图生图模式)
基于提供的参考图片,提取角色的面部五官特征、发型、体型和服装款式作为参考。
## 画风优先级规则
画风由用户选择的风格指令决定,严格遵循风格指令进行生成。
参考图仅用于保持角色身份特征(五官、发型、体型、服装结构),不能覆盖用户指定画风。
仅在未提供风格指令时,才可参考原图画风。
## 生成规则
1. 忽略原图的具体色调和光线
2. 使用自然柔和的摄影棚灯光
3. 绘制正常美观的人体比例
4. 不要复制原图的画质、模糊、噪点或瑕疵
5. 生成的图像必须清晰锐利、细节丰富、专业品质
6. 角色表情应为自然平静的中性表情,目光正视镜头
## 缺失部位自动补齐
如果参考图是半身或部分身体,请根据服装风格和人物特征合理补全未露出的部位:
- **缺少下半身**:根据上衣风格推断并绘制匹配的裤装/裙装
- **缺少脚部**:根据整体穿搭风格添加合适的鞋款
- **缺少手部/手臂**:根据姿态合理补全
- 保持整体风格一致,确保补全的部分与可见部分协调统一