window-scenery-skill

KO | EN

gitlite — search

#python #java #python3 #arduino #golang #machine-learning #rust #html #flask #javascript #seismology #nodejs

window-scenery-skill

★ 10 Open GitHub ↗

Codex skill for generating travel scenery viewed from vehicle windows

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# 车窗里的风景 / Window Scenery Skill

[中文说明](README.zh-CN.md) · [Visual Manual](#visual-manual) · [Examples](#example-gallery)

`window-scenery` is a Codex skill for generating photorealistic travel images seen from inside transportation: car windows, train windows, ferry cabins, plane windows, cable-car glass, and off-road vehicle frames.

The core idea is simple:

- The window frame becomes the picture frame.
- The trip itself becomes the narrative.
- The location decides the scenery.
- The transportation decides the viewpoint.

It is designed for prompts such as:

```text
车窗里的风景：北京天安门
车窗美景：
1. 威尼斯
2. 马尔代夫
3. 张家界
4. 瓦迪拉姆
```

## Example Gallery

| City | Boat | Plane |
|---|---|---|
| ![重庆 · 洪崖洞](https://raw.githubusercontent.com/dacnay816y62-hub/window-scenery-skill/main/examples/chongqing-hongyadong-final-cn.jpg) | ![威尼斯 · 大运河](https://raw.githubusercontent.com/dacnay816y62-hub/window-scenery-skill/main/examples/venice-grand-canal-final-cn.jpg) | ![马尔代夫 · 环礁](https://raw.githubusercontent.com/dacnay816y62-hub/window-scenery-skill/main/examples/maldives-atoll-final-cn.jpg) |

| Train | Cable Car | Desert |
|---|---|---|
| ![班夫 · 露易丝湖](https://raw.githubusercontent.com/dacnay816y62-hub/window-scenery-skill/main/examples/banff-lake-louise-final-cn.jpg) | ![张家界 · 武陵源](https://raw.githubusercontent.com/dacnay816y62-hub/window-scenery-skill/main/examples/zhangjiajie-final-cn.jpg) | ![瓦迪拉姆](https://raw.githubusercontent.com/dacnay816y62-hub/window-scenery-skill/main/examples/wadi-rum-final-cn.jpg) |

## Visual Manual

The Chinese visual manual explains the concept, prompt logic, transport matching, and location badge system.

| Cover | Motion Note |
|---|---|
| ![Manual cover](https://raw.githubusercontent.com/dacnay816y62-hub/window-scenery-skill/main/docs/manual/manual-01-cover.png) | ![Window-position motion note](https://raw.githubusercontent.com/dacnay816y62-hub/window-scenery-skill/main/docs/manual/manual-02-window-position.gif) |

| Core Logic | Transport Matching | Location Badge |
|---|---|---|
| ![Core logic](https://raw.githubusercontent.com/dacnay816y62-hub/window-scenery-skill/main/docs/manual/manual-03-core-logic.png) | ![Transport matching](https://raw.githubusercontent.com/dacnay816y62-hub/window-scenery-skill/main/docs/manual/manual-04-transport-matching.png) | ![Location badge](https://raw.githubusercontent.com/dacnay816y62-hub/window-scenery-skill/main/docs/manual/manual-05-location-badge.png) |

## What This Skill Does

The skill helps Codex:

- Parse one or many destination names.
- Classify each place by travel landscape type.
- Choose a plausible transportation mode and window shape.
- Compile an image prompt for a no-text base image.
- Add a consistent lower-left location badge afterward.
- Keep batch outputs varied across transport, light, composition, and geography.

The skill intentionally avoids ordinary landscape postcards. Every result should feel like a real moment in transit.

## Output Style

Default output:

- Single image per location.
- Vertical 3:4 travel photograph.
- Photorealistic and lightly cinematic.
- Visible vehicle or cabin window frame.
- Exterior scenery as the subject.
- Small lower-left badge in the current user language.

Avoid:

- Collages.
- Poster layouts.
- Travel-advertising gloss.
- Extra titles or watermarks.
- Missing window frames.
- Oversized location labels.

## Installation

Clone or copy this folder into your Codex skills directory:

```powershell
git clone https://github.com/dacnay816y62-hub/window-scenery-skill.git `
  "$env:USERPROFILE\.codex\skills\window-scenery"
```

If your Codex skills live somewhere else, copy the repository folder there and keep the folder name as `window-scenery`.

The required skill file is:

```text
SKILL.md
```

Optional supporting files:

```text
agents/openai.yaml
references/style-guide.md
scripts/overlay_badge.py
```

## Usage

Minimal:

```text
车窗里的风景：北京天安门
```

Standard:

```text
车窗里的风景：
地点：挪威洛福滕
交通工具：自动
时间：自动
天气：自动
比例：3:4
地点标：开启
```

Precise:

```text
车窗里的风景：
地点：西藏林芝
交通工具：SUV
构图：侧窗掠过型
时间：秋日下午
天气：晴朗
重点：尼洋河、雪山、松林
坐标：29.6547°N 94.3623°E
比例：3:4
地点标：开启
```

Batch:

```text
车窗里的风景：
1. 东京
2. 班夫
3. 马尔代夫
4. 威尼斯
5. 卡帕多奇亚
```

Batch rule: each numbered location must become one separate image. Do not merge destinations into one collage.

## Location Badge Workflow

Image models are often unreliable at rendering small text. This skill therefore recommends a two-stage workflow:

1. Generate a no-text base image.
2. Add the lower-left location badge programmatically.

Use the bundled script:

```powershell
python ".\scripts\overlay_badge.py" `
  --input "D:\Codex_Outputs\images\base.png" `
  --output "D:\Codex_Outputs\images\final.jpg" `
  --place "北京 · 天安门" `
  --coords "39.9087°N   116.3975°E"
```

The badge uses the current user language for the place label. For Chinese requests, use Chinese labels such as `北京 · 天安门`, not English-only labels.

## Repository Structure

```text
window-scenery-skill/
├── SKILL.md
├── README.md
├── agents/
│   └── openai.yaml
├── examples/
│   ├── banff-lake-louise-final-cn.jpg
│   ├── chongqing-hongyadong-final-cn.jpg
│   ├── maldives-atoll-final-cn.jpg
│   ├── venice-grand-canal-final-cn.jpg
│   ├── wadi-rum-final-cn.jpg
│   └── zhangjiajie-final-cn.jpg
├── docs/
│   └── manual/
│       ├── manual-01-cover.png
│       ├── manual-02-window-position.gif
│       ├── manual-03-core-logic.png
│       ├── manual-04-transport-matching.png
│       └── manual-05-location-badge.png
├── references/
│   └── style-guide.md
└── scripts/
    └── overlay_badge.py
```

## Notes

- `SKILL.md` is the agent-facing instruction file.
- `references/style-guide.md` stores detailed style, transport, composition, and test guidance.
- `scripts/overlay_badge.py` is a deterministic post-processing helper for the lower-left location badge.
- Example images are demonstration outputs and are not required for the skill to run.

window-scenery-skill

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README