图像到图像的身份保留插图
Author:Gagan Singh
2026/01/05 09:12
Description
创建全新插画场景,保留参考人物面部身份,改变姿态、背景与整体构图。
Tags
画像生成ビジュアル化コンテンツ生成
Content
```json
{
"input_type": "image_to_image",
"target_image": "{argument name=\"target image\" default=\"UPLOAD_IMAGE_HERE\"}",
"prompt": "创建一个全新的插画场景,仅将目标图像用作面部身份参考。精确保留相同的面部、肤色和身份,但改变其他所有元素。生成一个新的姿势、不同的身体姿态、新的摄像机角度和全新的构图。将背景完全替换为您选择的新地点和环境,例如时尚的城市街道、自然景观、咖啡馆、屋顶或艺术场景。采用现代数字插画风格,具有清晰的线条、平滑的阴影、柔和的渐变和精致的艺术外观(非绘画、非水彩)。服装、颜色、情绪和光线应进行创意重新设计,同时保持面部可识别且自然。最终结果应看起来像一幅全新的插画作品,而非原始图像的修改版本。",
"identity_strength": 0.9,
"style_strength": 0.8,
"allow_pose_change": true,
"allow_background_change": true,
"allow_scene_change": true,
"preserve_face_only": true,
"quality": "4K",
"negative_prompt": "相同姿势, 相同背景, 相同位置, 绘画纹理, 水彩, 油画, 动漫, 卡通夸张, 面部扭曲, 多余肢体, 模糊, 低细节"
}
```
---
**Original English:**
{
"input_type": "image_to_image",
"target_image": "{argument name="target image" default="UPLOAD_IMAGE_HERE"}",
"prompt": "Create a completely new illustrated scene using the target image only for facial identity reference. Preserve the same face, skin tone, and identity accurately, but change everything else. Generate a new pose, different body posture, new camera angle, and a fresh composition. Replace the background entirely with a new location and environment of your choice, such as an aesthetic urban street, nature landscape, café, rooftop, or artistic setting. Apply a modern digital illustration style with clean lines, smooth shading, soft gradients, and a polished artistic look (not painting, not watercolor). Clothing, colors, mood, and lighting should be redesigned creatively while keeping the face recognizable and natural. The final result should look like a newly illustrated artwork, not a modified version of the original image.",
"identity_strength": 0.9,
"style_strength": 0.8,
"allow_pose_change": true,
"allow_background_change": true,
"allow_scene_change": true,
"preserve_face_only": true,
"quality": "4K",
"negative_prompt": "same pose, same background, same location, painting texture, watercolor, oil paint, anime, cartoon exaggeration, distorted face, extra limbs, blur, low detail"
}