# 用 GPT-image2 一键出监控美学大片，1 个 skill 7 种风格我替你整理好了
**作者**: 神奇小喷菇AIGC
**日期**: 2026-04-19T09:53:02.000Z
**来源**: [https://x.com/xpg0970/status/2045802847573520723](https://x.com/xpg0970/status/2045802847573520723)
---
监控美学 (Surveillance Aesthetic) AI 画图技能（skill）
监控美学，也称「算法凝视风」(Algorithmic Gaze)，是一种将日常场景通过监控/AI视觉系统的视角重新呈现的视觉艺术风格。核心手法是在画面中叠加线框、检测框、数据标注等元素，营造一种「被凝视」「被捕捉」「被分析」的技术氛围。
使用场景
当用户需要以下任务时激活本技能：
1.生成监控美学 / 算法凝视风格的 AI 图像
2.为照片或场景添加监控摄像头视角效果
3.制作带有目标检测框、数据标注的艺术图像
4.创作赛博朋克 / 反乌托邦 / 技术凝视风格的视觉作品
5.任何涉及「监控」「凝视」「检测」「算法」「CCTV」美学关键词的画图请求
视觉语言核心要素
监控美学的视觉辨识度来自以下 六大核心元素 的组合运用：
1. 检测框与线框 (Bounding Box & Wireframe)
最核心的标志性元素。 矩形框选画面中的主体（人物、物体、动物），模拟 AI 目标检测的视觉输出。
**最核心的标志性元素。** 矩形框选画面中的主体（人物、物体、动物），模拟 AI 目标检测的视觉输出。
| 属性 | 规范 |
|------|------|
| 框体样式 | 细线矩形，1-2px，直角或微圆角 |
| 颜色 | 绿色 `#00FF00`、青色 `#00FFFF`、红色 `#FF0000`、黄色 `#FFFF00` |
| 标注文字 | 框旁附类别标签，如 `person 0.97`、`cat 0.85`、`vehicle 0.92` |
| 置信度 | 标签常带 0.xx 格式的置信度数值 |
| 数量 | 画面中通常有 2-5 个检测框，主体框最大最醒目 |
### 2. HUD 数据叠层 (Heads-Up Display Overlay)
模拟监控系统界面的信息叠加层。
| 元素 | 位置 | 格式示例 |
|------|------|----------|
| 时间戳 | 左下或右下角 | `2024-03-15 14:32:07` |
| 摄像头编号 | 左上角 | `CAM-07` / `NODE_12` / `UNIT-A3` |
| 录制状态 | 右上角 | `● REC` (红点 + REC) |
| 坐标信息 | 底部 | `LAT 31.2304 LON 121.4737` |
| 帧率/分辨率 | 角落 | `30fps 1920x1080` / `H.264` |
### 3. 色调与质感 (Color Grading & Texture)
| 子风格 | 色调特征 | 质感特征 |
|--------|----------|----------|
| 经典 CCTV | 低饱和、偏冷灰绿 | 轻微噪点、压缩伪影 |
| 夜视模式 | 单色绿 `#00FF00` 基调 | 高噪点、光晕溢出 |
| 热成像 | 红橙黄蓝渐变伪彩色 | 低分辨率模糊、热力分布 |
| 高清监控 | 冷调、高对比度 | 清晰但略微失真的广角畸变 |
| 算法分析 | 暗底 + 霓虹标注 | 干净底图 + 高亮数据层 |
### 4. 网格与辅助线 (Grid & Guide Lines)
- 透视网格线（地面延伸的虚线网格）
- 面部关键点连线（三角网格覆盖人脸）
- 运动轨迹线（虚线箭头标注运动方向）
- 距离标尺（标注目标间距离）
### 5. 镜头特征 (Lens Characteristics)
| 特征 | 表现 |
|------|------|
| 广角畸变 | 鱼眼或桶形畸变，边缘弯曲 |
| 景深 | 通常全景深，所有物体等清晰度（监控镜头特征） |
| 视角 | 高位俯瞰（天花板视角）或平视固定机位 |
| 暗角 | 画面四角轻微暗角渐变 |
### 6. 文字与数据标注 (Text & Data Annotation)
- 字体：等宽字体（Courier、Consolas 风格）
- 颜色：白色、绿色或青色，半透明底色衬托
- 内容：目标ID、行为分类、区域标签、系统状态
---
## 七大子风格与 Prompt 模板
### 风格 1：经典 CCTV 监控
> 模拟传统安防摄像头画面，低画质、固定机位、时间戳水印。
**视觉关键词：** 低分辨率、灰绿色调、噪点、鱼眼畸变、时间戳、CAM编号
**Prompt 结构：**
```
CCTV security camera footage of [场景描述], low resolution, grainy surveillance camera quality,
slight fisheye lens distortion, desaturated cold green tint, timestamp overlay "YYYY-MM-DD HH:MM:SS"
in bottom corner, camera ID "CAM-XX" in top left, REC indicator in top right,
dark vignette edges, fixed angle overhead view, [补充描述]
```
**场景示例：**
```
CCTV security camera footage of a person walking alone through an empty convenience store at night,
low resolution, grainy surveillance camera quality, slight fisheye lens distortion,
desaturated cold green tint, timestamp overlay "2024-11-23 03:42:17" in bottom left corner,
camera ID "CAM-04" in top left, red dot REC indicator top right, fluorescent lighting,
dark vignette edges, overhead angle
```
---
### 风格 2：AI 目标检测 (Object Detection)
> 画面中的主体被绿色/彩色矩形框标注，附带类别标签和置信度，模拟 YOLO/Faster R-CNN 等算法输出。
**视觉关键词：** 检测框、分类标签、置信度分数、暗色底图、霓虹标注
**Prompt 结构：**
```
AI computer vision object detection visualization of [场景描述],
rectangular bounding boxes around detected objects with class labels and confidence scores,
green and cyan wireframe rectangles labeled "person 0.97" "car 0.89" "[类别] [分数]",
dark moody atmosphere, technical HUD overlay, clean monospace text annotations,
slight digital processing look, [补充描述]
```
**场景示例：**
```
AI computer vision object detection visualization of a busy Tokyo street crossing at night,
multiple rectangular bounding boxes around pedestrians labeled "person 0.95" "person 0.91",
around vehicles labeled "car 0.88" "bus 0.93", green and cyan wireframe detection rectangles,
neon-lit urban background, technical HUD overlay with frame counter,
rain-wet reflections on asphalt, cinematic atmosphere
```
---
### 风格 3：夜视模式 (Night Vision)
> 单色绿基调，高噪点，红外光斑，模拟军用/安防夜视设备画面。
**视觉关键词：** 单色绿、高噪点、红外光晕、暗夜场景、军事感
**Prompt 结构：**
```
Night vision camera view of [场景描述], monochrome green tint,
high ISO noise and grain, infrared illumination glow on subjects,
green phosphor display aesthetic, bright spots blooming with light halos,
military surveillance style, generation III night vision look,
vignette darkening at edges, [补充描述]
```
---
### 风格 4：热成像 (Thermal Imaging)
> 伪彩色热力图，人体和热源呈红橙黄高亮，背景为蓝紫冷色，模拟红外热成像仪输出。
**视觉关键词：** 伪彩色、热力分布、红橙黄蓝、低分辨率、温度标尺
**Prompt 结构：**
```
Thermal infrared camera view of [场景描述], false color heat map visualization,
human bodies glowing in red-orange-yellow against blue-purple cool background,
FLIR thermal imaging style, temperature scale bar on side,
low resolution blocky pixels, heat signatures radiating from warm objects,
technical readout overlay with temperature values, [补充描述]
```
---
### 风格 5：无人机俯瞰 (Drone Surveillance)
> 高空鸟瞰视角，目标被圆形或方形追踪框锁定，模拟军用/警用无人机画面。
**视觉关键词：** 鸟瞰视角、追踪框、十字准星、灰调、坐标信息
**Prompt 结构：**
```
Aerial drone surveillance footage of [场景描述], top-down bird's eye view,
high altitude perspective, tracking crosshair locked on target subject,
square tracking bracket around subject, grayscale with slight blue tint,
GPS coordinates overlay, altitude and speed readout,
military UAV camera aesthetic, slight atmospheric haze, [补充描述]
```
---
### 风格 6：面部识别分析 (Facial Recognition)
> 人脸上叠加关键点网格、特征向量可视化、身份匹配信息，模拟人脸识别系统界面。
**视觉关键词：** 面部网格、关键点、特征向量、身份信息、匹配度
**Prompt 结构：**
```
Facial recognition system analysis view of [人物描述],
geometric mesh wireframe overlay on face mapping facial landmarks,
green dotted lines connecting facial keypoints, identity matching panel on side,
biometric data readout, confidence percentage, age and expression analysis labels,
dark interface background, technical blue-green accent lighting, [补充描述]
```
---
### 风格 7：数据标注美学 (Data Annotation Aesthetic)
> 最具当代艺术感的子风格。画面整洁优美，但被精心设计的标注框和数据层赋予技术凝视感。适合用于社交媒体和艺术创作。
**视觉关键词：** 极简标注、高质感底图、精致线框、艺术化数据层
**Prompt 结构：**
```
Beautiful [场景描述] with elegant data annotation overlay,
minimalist thin white bounding boxes around key subjects,
small clean sans-serif labels with classification text,
high quality photograph underneath with subtle cool color grading,
artistic surveillance aesthetic, the contrast between beauty and clinical annotation,
Instagram-worthy composition, [补充描述]
```
**场景示例：**
```
Beautiful golden hour portrait of a young woman in a flower field with elegant data annotation overlay,
minimalist thin white bounding boxes around her face labeled "face_01" and hands labeled "hand_L" "hand_R",
small bounding box around a butterfly labeled "lepidoptera 0.73",
high quality photograph with subtle cool desaturated color grading,
artistic surveillance aesthetic, the contrast between natural beauty and clinical data annotation,
soft bokeh background, gentle warm light
```
---
## 工作流程
### 第 1 步：理解用户需求
向用户确认以下信息：
```
场景描述：[想要呈现的画面内容]
子风格偏好：[经典CCTV / AI检测 / 夜视 / 热成像 / 无人机 / 面部识别 / 数据标注]
（默认：AI 目标检测 + 数据标注混合风格）
画面氛围：[冷峻科技 / 反乌托邦 / 艺术唯美 / 紧张悬疑 / 日常冷幽默]
尺寸比例：[1:1 正方形 / 16:9 宽屏 / 3:2 标准 / 9:16 竖屏]
补充要求：[其他特殊需求]
```
如果用户只提供了简单描述（如"给我画一张监控风的猫"），不需要逐项确认，直接基于描述选择最合适的子风格和参数进行创作。
### 第 2 步：构建 Prompt
根据用户需求，按以下结构组装最终 Prompt：
```
[子风格模板基础句] + [用户场景描述] + [核心视觉元素] + [氛围/光照] + [补充细节]
```
**Prompt 组装原则：**
- 英文撰写 Prompt（AI 画图模型对英文理解更准确）
- 将监控元素关键词放在 Prompt 前半部分以增强权重
- 场景描述要具体（时间、地点、人物动作、环境细节）
- 标注内容要合理（检测框的类别和置信度要符合场景逻辑）
- 包含至少 2-3 个 HUD 数据元素（时间戳、摄像头编号、REC 标识等）
### 第 3 步：选择工具与生成
| 需求场景 | 推荐工具 | 原因 |
|----------|----------|------|
| 快速出图 / 批量风格探索 | `generate-image-fast` | 速度快，适合迭代 |
| 高品质最终作品 | `generate-image-hq` | OpenAI 模型，细节更好 |
| 基于现有照片添加监控风格 | `generate-image-hq` + `referenceImageUrls` | 支持图片编辑 |
**尺寸建议：**
| 用途 | 推荐尺寸 | 工具参数 |
|------|----------|----------|
| 社交媒体帖子 | 1024x1024 | `size: "1:1"` 或 `width: 1024, height: 1024` |
| 电影感横幅 | 1536x1024 | `size: "3:2"` 或 `width: 1536, height: 1024` |
| 超宽屏 / 桌面壁纸 | 1920x1080 | `width: 1920, height: 1080` |
| 手机壁纸 / 短视频封面 | 1024x1536 | `size: "1024x1536"` 或 `width: 1024, height: 1536` |
### 第 4 步：输出与优化
1. 生成图像后展示给用户
2. 如果效果不理想，调整以下参数：
- **检测框不够明显** → 在 Prompt 中强调 `prominent bounding boxes`, `clearly visible detection rectangles`
- **监控感不够强** → 增加 HUD 元素描述，添加 `grainy`, `low quality camera`, `CCTV aesthetic`
- **画面太暗/太亮** → 调整光照描述，如 `dim fluorescent lighting`, `harsh overhead light`
- **风格太写实/太抽象** → 在 `photorealistic` 和 `digital art` 之间调整
3. 支持用户在多个子风格间切换对比
---
## Prompt 增强词库
### 监控氛围增强词
| 类别 | 增强关键词 |
|------|-----------|
| 画质 | `grainy`, `low resolution`, `compression artifacts`, `interlaced`, `pixelated` |
| 镜头 | `fisheye distortion`, `wide angle`, `fixed camera`, `barrel distortion`, `overhead mounted camera` |
| 光照 | `fluorescent lighting`, `harsh overhead light`, `infrared illumination`, `sodium vapor light` |
| 色调 | `desaturated`, `cold blue tint`, `green monochrome`, `washed out colors`, `low contrast` |
| 氛围 | `eerie`, `voyeuristic`, `clinical`, `impersonal`, `panoptic`, `liminal space` |
| 技术 | `HUD overlay`, `data visualization`, `algorithmic`, `computer vision output`, `machine learning detection` |
### 标注文本素材库
**目标检测类别：**
```
person, pedestrian, vehicle, car, bicycle, motorcycle, bus, truck,
cat, dog, bird, backpack, handbag, umbrella, chair, table,
bottle, cup, phone, laptop, book, plant, traffic_light, stop_sign
```
**系统状态文本：**
```
● REC | RECORDING | LIVE | PLAYBACK
CAM-01 ~ CAM-99 | NODE_A1 ~ NODE_F8
ZONE-ALPHA | SECTOR-7 | PERIMETER-NORTH
ALERT | WARNING | TRACKING | SCANNING
MOTION DETECTED | ANOMALY DETECTED | TARGET ACQUIRED
```
**数据格式：**
```
时间戳：2024-11-23 03:42:17.034
坐标：LAT 31.2304°N LON 121.4737°E
帧率：30fps / 60fps
编码：H.264 / H.265
分辨率：1920x1080 / 2560x1440 / 3840x2160
```
---
## 创作灵感：场景 x 风格 推荐组合
| 场景 | 推荐子风格 | 氛围方向 |
|------|-----------|----------|
| 城市街头行人 | AI 目标检测 | 冷峻科技、都市异化 |
| 便利店/超市 | 经典 CCTV | 日常冷幽默、深夜孤独感 |
| 自然风景/动物 | 数据标注美学 | 唯美与技术的反差 |
| 人像/自拍 | 面部识别分析 | 身份焦虑、隐私议题 |
| 城市鸟瞰/建筑群 | 无人机俯瞰 | 全景监控、权力凝视 |
| 夜间室外 | 夜视模式 | 军事感、紧张悬疑 |
| 人群聚集 | 热成像 | 公共卫生、安防预警 |
| 花园/咖啡厅 | 数据标注美学 | 日常美好被算法解构 |
| 宠物日常 | AI 目标检测 | 反差萌、科技幽默 |
---
## 批量创作模式
当用户需要同一场景的多风格对比或系列创作时，使用 `generate-image-fast` 批量生成：
```json
{
"dispatch": "generate-image-fast",
"args": {
"images": [
{ "prompt": "[风格1 完整 Prompt]", "width": 1024, "height": 1024 },
{ "prompt": "[风格2 完整 Prompt]", "width": 1024, "height": 1024 },
{ "prompt": "[风格3 完整 Prompt]", "width": 1024, "height": 1024 }
]
}
```
---
## 将现有照片转为监控美学风格
使用 `generate-image-hq` 的图片编辑能力：
```json
{
"dispatch": "generate-image-hq",
"args": {
"images": [
{
"prompt": "Transform this photo into surveillance camera aesthetic: add green bounding boxes around main subjects with detection labels and confidence scores, add timestamp overlay, camera ID, REC indicator, apply slight desaturation and cool tint, add subtle scan lines and noise",
"referenceImageUrls": ["用户提供的图片URL"],
"size": "1024x1024"
}
]
}
```
---
## 注意事项
1. **隐私与伦理**：监控美学是一种艺术表达形式，创作时应注意：
- 不生成可辨识真实个人的监控画面
- 不模拟真实安防系统界面用于欺骗
- 创作目的是艺术探讨与审美表达
2. **Prompt 语言**：始终使用英文撰写画图 Prompt，但与用户的交流使用中文
3. **风格平衡**：监控元素（线框、标注）与画面美感之间要取得平衡，避免标注元素过多导致画面杂乱
4. **迭代优化**：AI 画图对监控 HUD 元素的还原度有限，可能需要多次调整 Prompt 或后期叠加。如果单次生成效果不理想，尝试：
- 简化 HUD 元素数量（保留最核心的 2-3 个）
- 将「检测框」作为最高优先级元素强调
- 使用更具体的视觉描述替代抽象概念
（完）
示例：
1) 经典 CCTV 监控（Classic CCTV）
CCTV security camera footage of a small apartment hallway at night, one person carrying a bag walking past the camera, low resolution, grainy surveillance camera quality, slight fisheye lens distortion, desaturated cold green tint, subtle scan lines and compression artifacts, timestamp overlay "2026-04-19 01:08:23" in bottom left, camera ID "CAM-07" in top left, red dot "REC" indicator in top right, harsh overhead fluorescent lighting, dark vignette edges, fixed ceiling-mounted angle, impersonal voyeuristic mood, monospace overlay text
2) AI 目标检测（Object Detection Visualization）
AI computer vision object detection visualization of a busy city intersection in light rain, multiple rectangular bounding boxes around pedestrians labeled "person 0.96" "person 0.92" "person 0.87", bounding boxes around vehicles labeled "car 0.90" "bus 0.93" "motorcycle 0.84", green and cyan thin wireframe detection rectangles, clean monospace text annotations, technical HUD overlay with "FRAME 45218" "30fps" "H.264", timestamp "2026-04-19 21:17:05" bottom right, slight digital processing look, neon reflections on wet asphalt, clinical algorithmic gaze
3) 夜视模式（Night Vision）
Night vision camera view of a quiet backyard at midnight, a person standing near a fence and a dog moving in the grass, monochrome green tint, high ISO noise and heavy grain, infrared illumination glow on subjects, green phosphor display aesthetic, bright spots blooming with light halos, military surveillance style, vignette darkening at edges, subtle scan lines, timestamp "2026-04-19 00:42:11" and camera label "NV-CAM-02" in monospace text
4) 热成像（Thermal / FLIR）
Thermal infrared camera view of a subway platform, several human bodies glowing in red-orange-yellow against a blue-purple cool background, FLIR thermal imaging style, false color heat map visualization, temperature scale bar on the right, low resolution blocky pixels, heat signatures radiating from warm objects, technical readout overlay showing "MAX 36.8C MIN 18.2C", timestamp "2026-04-19 18:05:33" and device label "FLIR UNIT-3" in monospace text
5) 无人机俯瞰（Drone Surveillance / UAV）
Aerial drone surveillance footage of a rooftop parking lot at dusk, top-down bird's eye view, one moving vehicle highlighted with a square tracking bracket and crosshair locked on target, grayscale with slight blue tint, GPS coordinates overlay "LAT 31.2304N LON 121.4737E", altitude readout "ALT 320m" and speed "SPD 41km/h", HUD text "TRACKING" "TARGET ACQUIRED", slight atmospheric haze, vignette edges, timestamp "2026-04-19 19:28:33"
6) 面部识别分析（Facial Recognition Analysis）
Facial recognition system analysis view of a person sitting in a coffee shop by the window, geometric mesh wireframe overlay mapping facial landmarks, green dotted lines connecting keypoints, a bounding box labeled "face 0.98", identity matching panel on the right with "MATCH 0.86" "ID SUBJECT_014" "AGE 26" "EMOTION NEUTRAL", dark UI overlay with blue-green accents, subtle noise and scan lines, timestamp "2026-04-19 14:21:09" and camera ID "NODE_12" in monospace text, impersonal surveillance aesthetic
7) 数据标注美学（Data Annotation Aesthetic / Artistic）
Beautiful golden hour street portrait of a person walking past a flower shop, elegant data annotation overlay with minimalist thin white bounding boxes around the face labeled "face_01" and hands labeled "hand_L" "hand_R", a small box around a bicycle labeled "bicycle 0.78", clean sans-serif labels, subtle cool desaturated color grading, artistic surveillance aesthetic (beauty contrasted with clinical annotation), high quality photograph underneath, small timestamp "2026-04-19 17:36:12" in a corner
## 相关链接
- [神奇小喷菇AIGC](https://x.com/xpg0970)
- [@xpg0970](https://x.com/xpg0970)
- [392](https://x.com/xpg0970/status/2045802847573520723/analytics)
- [#00FF00](https://x.com/search?q=%2300FF00&src=hashtag_click)
- [#00FFFF](https://x.com/search?q=%2300FFFF&src=hashtag_click)
- [#FF0000](https://x.com/search?q=%23FF0000&src=hashtag_click)
- [#FFFF00](https://x.com/search?q=%23FFFF00&src=hashtag_click)
- [#00FF00](https://x.com/search?q=%2300FF00&src=hashtag_click)
- [Upgrade to Premium](https://x.com/i/premium_sign_up)
- [5:53 PM · Apr 19, 2026](https://x.com/xpg0970/status/2045802847573520723)
- [392 Views](https://x.com/xpg0970/status/2045802847573520723/analytics)
---
*导出时间: 2026/4/19 20:18:49*