\r \r

Image-2 Skill\r

\r

Create, edit, transform, and analyze images with GPT-4o's native image generation API\r \r

When to Use This Skill\r

\r Use this skill whenever the user needs to:\r

• Generate images from text descriptions ("画一张...", "生成图片...", "create an image of...")\r
• Edit existing images with natural language ("把背景去掉", "add a sunset", "换成蓝色")\r
• Create variations of an image ("生成几个变体", "make 4 variations")\r
• Analyze/describe images ("这张图是什么", "describe this image", "提取文字")\r
• Remove backgrounds ("去除背景", "remove background")\r
• Style transfer ("变成水彩风格", "make it look like Van Gogh")\r
• Create marketing visuals ("设计海报", "make a social media post")\r
• Product photography ("产品图", "product shot on white background")\r
• UI/UX mockups ("界面设计", "app mockup", "website screenshot")\r \r

Core Workflows\r

\r

Workflow 1: Text-to-Image Generation\r

\r When the user describes an image they want to create:\r \r

1. Enhance the prompt — Automatically add quality boosters:\r
   • Append professional photography/art terms based on context\r
   • Add lighting, composition, and mood details if not specified\r
   • Specify output format and dimensions if needed\r \r
2. Call the API — Use generateImage() with the enhanced prompt:\r

   const result = await generateImage(enhancedPrompt, { size, quality, style });\r
   ```\r
   

\r 3. Save and present — Download the image to the project directory and show the user:\r

• Save to ./generated-images/ by default\r
• Return the file path and a brief description\r \r

Workflow 2: Image Editing\r

\r When the user wants to modify an existing image:\r \r

1. Locate the source image — Find the image file path from the conversation context\r
2. Parse the edit intent — Understand what changes the user wants\r
3. Call the edit API — Use editImage() with the source and instruction:\r

   const result = await editImage(imagePath, editInstruction, { mask: maskPath });\r
   ```\r
   

4. Present the result — Show the edited image and describe what changed\r \r

Workflow 3: Image Analysis\r

\r When the user asks about an image:\r \r

1. Get the image — From file path or URL\r
2. Analyze with GPT-4o Vision — Use describeImage():\r

   const result = await describeImage(imageSource, question);\r
   ```\r
   

3. Report findings — Present the analysis in a structured format\r \r

Workflow 4: Batch Generation\r

\r When the user needs multiple images:\r \r

1. Parse the batch request — Understand variations needed\r
2. Generate in parallel — Call generateImage() for each variant\r
3. Organize results — Save with descriptive filenames\r \r

Prompt Enhancement Rules\r

\r When generating images, automatically enhance the user's prompt:\r \r

Quality Boosters (always append unless user specifies quality)\r

professional quality, high resolution, sharp details\r
```\r
\r
### Context-Based Additions\r
| User Intent | Auto-Add |\r
|-------------|----------|\r
| Product photo | "studio lighting, clean background, commercial photography" |\r
| Portrait | "professional portrait photography, natural lighting" |\r
| Social media | "eye-catching, vibrant colors, modern design" |\r
| Illustration | "detailed illustration, professional artist quality" |\r
| Logo/branding | "clean vector style, scalable, minimal details" |\r
| Architecture | "architectural visualization, realistic rendering" |\r
| Food | "appetizing, food styling, professional food photography" |\r
| UI mockup | "clean design, modern interface, pixel-perfect" |\r
\r
### Size Recommendations\r
| Use Case | Recommended Size |\r
|----------|-----------------|\r
| Social media post | `1024x1024` (square) |\r
| Story/vertical | `1024x1792` |\r
| Banner/landscape | `1792x1024` |\r
| Product listing | `1024x1024` |\r
| Presentation | `1792x1024` |\r
| Wallpaper | `1792x1024` |\r
\r
## Style Presets\r
\r
Quick style references for common requests:\r
\r
| Preset Name | Style Description |\r
|-------------|-------------------|\r
| `product` | Clean white background, studio lighting, commercial photography |\r
| `lifestyle` | Natural setting, warm lighting, aspirational mood |\r
| `minimalist` | Simple composition, negative space, clean lines |\r
| `vintage` | Retro color grading, film grain, nostalgic mood |\r
| `futuristic` | Neon accents, dark background, sci-fi aesthetic |\r
| `watercolor` | Soft edges, pastel palette, artistic brush strokes |\r
| `3d-render` | Octane render, realistic materials, dramatic lighting |\r
| `anime` | Japanese animation style, vibrant, expressive |\r
| `sketch` | Pencil drawing, hand-drawn, artistic |\r
| `flat-design` | Vector style, bold colors, geometric shapes |\r
\r
## API Reference\r
\r
### `generateImage(prompt, options)`\r
Generate a new image from text description.\r
\r
**Parameters:**\r
- `prompt` (string) — Image description (auto-enhanced by this skill)\r
- `options` (object):\r
  - `size` — `1024x1024` | `1024x1792` | `1792x1024` (default: `1024x1024`)\r
  - `quality` — `standard` | `hd` (default: `standard`)\r
  - `style` — `vivid` | `natural` (default: `vivid`)\r
  - `model` — `gpt-image-2` | `dall-e-3` (default: `gpt-image-2`)\r
  - `saveTo` — File path to save the image (default: `./generated-images/`)\r
\r
**Returns:** `{ success, url, localPath, revisedPrompt }`\r
\r
### `editImage(imagePath, prompt, options)`\r
Edit an existing image with natural language instructions.\r
\r
**Parameters:**\r
- `imagePath` (string) — Path to the source image\r
- `prompt` (string) — Edit instruction\r
- `options` (object):\r
  - `mask` — Path to mask image (white = edit area, black = keep)\r
  - `size` — Output size\r
  - `model` — `gpt-image-2` | `dall-e-3` (default: `gpt-image-2`)\r
\r
**Returns:** `{ success, url, localPath }`\r
\r
### `generateVariations(imagePath, options)`\r
Generate creative variations of an existing image.\r
\r
**Parameters:**\r
- `imagePath` (string) — Path to the source image\r
- `options` (object):\r
  - `count` — Number of variations 1-4 (default: 2)\r
  - `size` — Output size\r
\r
**Returns:** `{ success, variations: [{ url, localPath }] }`\r
\r
### `describeImage(imageSource, question)`\r
Analyze an image using GPT-4o Vision.\r
\r
**Parameters:**\r
- `imageSource` (string) — File path or URL of the image\r
- `question` (string|null) — Specific question about the image (default: general description)\r
\r
**Returns:** `{ success, description }`\r
\r
### `downloadImage(url, savePath)`\r
Download a generated image to local storage.\r
\r
**Parameters:**\r
- `url` (string) — Image URL from generation API\r
- `savePath` (string|null) — Local file path (default: auto-generated in `./generated-images/`)\r
\r
**Returns:** `{ success, localPath }`\r
\r
## Error Handling\r
\r
| Error | Cause | Resolution |\r
|-------|-------|------------|\r
| `Invalid API key` | OPENAI_API_KEY not set or invalid | Check environment variable |\r
| `Content policy violation` | Prompt violates safety guidelines | Rephrase the prompt |\r
| `Rate limit exceeded` | Too many requests | Wait and retry with backoff |\r
| `Image too large` | Source image exceeds size limit | Resize to under 4MB |\r
| `Timeout` | Generation took too long | Simplify prompt or retry |\r
\r
## Best Practices\r
\r
1. **Always enhance prompts** — Don't pass raw user input directly to the API\r
2. **Save locally** — Download generated images; URLs expire after 1 hour\r
3. **Use appropriate sizes** — Match the output size to the use case\r
4. **Prefer gpt-image-2** — Better quality and text rendering than dall-e-3\r
5. **Batch thoughtfully** — Generate 2-4 images max per request to avoid rate limits\r
6. **Describe edits clearly** — Be specific about what to change and where\r
\r
## Changelog\r
\r
### v1.1.0\r
- Added GPT-4o native image generation support (gpt-image-2 model)\r
- Added automatic prompt enhancement workflow\r
- Added image download and local save functionality\r
- Added style presets for quick reference\r
- Added batch generation workflow\r
- Improved error handling and documentation\r
\r
### v1.0.0\r
- Initial release with DALL-E 3 support\r
- Basic generate, edit, variations, and describe functions\r
\r
---\r
\r
**Tags:** `image-generation` `AI-art` `GPT-4o` `image-2` `gpt-image-2` `visual-creation` `marketing` `product-photos` `illustration` `design` `openai` `dall-e` `image-editing` `background-removal` `style-transfer` `ui-mockup`\r