---

name: generating-images description: Gemini APIを使用して画像を生成し、リポジトリに保存。ランディングページ、ヒーローイメージ、プロダクト画像、アバター、ダミー画像、プレースホルダー画像が必要なときに使用。画像生成、AI画像、自動生成。

Generating Images

Generate images using Gemini API and save them to the repository.

Quick Start

Copy# Set API key
export GEMINI_API_KEY="your-api-key"

# Generate an image (uses flash model by default)
python ~/.claude/skills/generating-images/scripts/generate_image.py \
  --prompt "A modern minimalist hero image for a tech startup" \
  --output "./assets/images/hero.png"

Key Concepts

Models

Other Settings

• Output: PNG format
• Aspect ratios: 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3
• API Key: Environment variable GEMINI_API_KEY

Workflows

Workflow 1: Basic Generation (Start with Flash)

1. Understand user's image requirements (purpose, style, colors)
2. Craft an effective prompt using guidelines below
3. Generate with flash model first (fast iteration)
4. Show result to user

Copypython ~/.claude/skills/generating-images/scripts/generate_image.py \
  --prompt "Your descriptive prompt here" \
  --output "./path/to/image.png" \
  --aspect-ratio "16:9"

Workflow 2: Quality Upgrade (Switch to Pro)

If user requests regeneration or higher quality:

Copypython ~/.claude/skills/generating-images/scripts/generate_image.py \
  --prompt "Your descriptive prompt here" \
  --output "./path/to/image.png" \
  --quality pro

When to switch to Pro:

• User says "make it better", "higher quality", "more detailed"
• After 2+ regeneration attempts with flash
• For final production images
• When fine details matter (text, faces, complex scenes)

Workflow 3: Edit with Reference Image

Use an existing image as a starting point:

Copypython ~/.claude/skills/generating-images/scripts/generate_image.py \
  --prompt "Add a gradient overlay and modern typography" \
  --input "./reference.png" \
  --output "./assets/hero-with-text.png"

Workflow 4: Iterative Refinement

1. First attempt: Flash model with initial prompt
2. If unsatisfied: Refine prompt, try flash again
3. If still unsatisfied: Switch to pro model
4. Final output: Save to appropriate location

Prompt Guidelines

Core principle: "Describe the scene, don't just list keywords."

Copy❌ Bad:  "hero image, blue, tech, modern"
✓ Good: "A modern hero image for a tech startup, featuring flowing
         abstract 3D waves in electric blue gradients with soft lighting"

Prompt Structure

CopyA [style] [type] of [subject], [action/state], set in [environment].
The [lighting] creates [mood]. [Technical details].

Quick Templates

Key Techniques

For photorealism: Use photography terms

• "85mm lens", "shallow depth of field", "three-point lighting", "golden hour"

For illustrations: Specify style explicitly

• "flat design", "minimalist", "3D rendered", "vector art", "isometric"

For quality: Add descriptors

• "commercial quality", "8K resolution", "professional", "high-end"

Detailed prompt guide: See PROMPTS.md for comprehensive examples and techniques.

Configuration

Environment Variable

Script Options

Error Handling

API key not set:

CopyError: GEMINI_API_KEY environment variable is not set
→ Set it: export GEMINI_API_KEY='your-api-key'

Rate limit exceeded:

CopyError: Rate limit exceeded
→ Wait a few seconds and retry

Content blocked by safety filter:

CopyError: No image data in response
→ Modify prompt to avoid potentially sensitive content

Examples

Example 1: Landing Page Hero

Copy# First try with flash
python ~/.claude/skills/generating-images/scripts/generate_image.py \
  --prompt "A stunning hero image for a modern tech startup..." \
  --output "./public/images/hero.png" \
  --aspect-ratio "16:9"

# If quality insufficient, retry with pro
python ~/.claude/skills/generating-images/scripts/generate_image.py \
  --prompt "A stunning hero image for a modern tech startup..." \
  --output "./public/images/hero.png" \
  --aspect-ratio "16:9" \
  --quality pro

Example 2: Quick Placeholder

Copypython ~/.claude/skills/generating-images/scripts/generate_image.py \
  --prompt "A simple placeholder image with subtle gray geometric pattern" \
  --output "./assets/placeholder.png"

Integration

After generating, add to your document:

Copy![Hero Image](./assets/images/hero.png)

Dependencies

Copypip install google-genai Pillow

Resources

• Gemini API Image Generation Docs