Nano Banana Pro Image Generator

Example Transformations

User says: "Make an app icon for a notes app"

You generate:

Minimalist 3D app icon design, floating notebook with softly curled pages, subtle pencil resting diagonally,
soft gradient background from warm cream to light amber, gentle ambient occlusion shadows,
rounded superellipse container shape, iOS/macOS design language, centered composition,
studio lighting with soft diffusion, clean vector-like render quality

User says: "Hero image for AI startup landing page"

You generate:

Abstract visualization of neural network, flowing geometric nodes connected by luminescent data streams,
deep space blue to electric cyan gradient background, volumetric god rays piercing through digital clouds,
wide 16:9 cinematic composition, dramatic three-point lighting with cyan key and purple fill,
futuristic tech aesthetic, photorealistic 3D render quality, depth of field with sharp foreground elements

Style Transfer Protocol

When the user provides a reference image:

1. Analyze the Reference

   • Identify dominant colors and palette
   • Detect artistic technique (photography, illustration, 3D, etc.)
   • Note composition patterns and framing style
   • Observe lighting characteristics
   • Identify texture and detail level

2. Extract Style Markers

   • Art movement or genre
   • Color temperature (warm/cool)
   • Contrast level
   • Edge treatment (sharp/soft)
   • Surface quality (matte/glossy/textured)

3. Construct Style-Matched Prompt Add preservation directives:

   Create [new subject] in the exact artistic style of the reference image.
   Preserve: color palette ({extracted colors}), {technique} style,
   {lighting type} lighting, {mood} atmosphere.
   Match the reference's {specific style elements}.
   

Using the Generator Script

The generation script is located at:

${CLAUDE_PLUGIN_ROOT}/scripts/generate.py

Basic Generation

python3 "${CLAUDE_PLUGIN_ROOT}/scripts/generate.py" "your optimized prompt here"

With Options

# Specific aspect ratio
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/generate.py" "prompt" --aspect-ratio 16:9

# High resolution
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/generate.py" "prompt" --resolution 4K

# Style transfer with reference
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/generate.py" "prompt" --reference /path/to/ref.png

# Custom output path
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/generate.py" "prompt" -o /path/to/output.png

Available Options

• --provider, -p: API provider (google or openrouter)
• --aspect-ratio, -a: 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9
• --resolution, -s: 1K, 2K (default), 4K
• --reference, -r: Path(s) to reference image(s) for style transfer
• --output, -o: Output file path
• --output-dir, -d: Output directory

Provider Selection

# Use Google API (default)
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/generate.py" "prompt" --provider google

# Use OpenRouter API
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/generate.py" "prompt" --provider openrouter

Workflow

Standard Generation Flow

1. Understand the Request

   • What type of image? (icon, banner, illustration, photo, etc.)
   • What style? (minimalist, realistic, artistic, etc.)
   • What purpose? (app, website, marketing, etc.)
   • What dimensions needed?

2. Generate Optimized Prompt

   • Apply the prompt formula
   • Include all relevant details
   • Add technical specifications
   • Consider the use case context

3. Execute Generation

   • Call the generate.py script with appropriate parameters
   • Use suitable aspect ratio for the use case

4. Present Result

   • Show the generated image to the user
   • Explain what was generated
   • Offer refinement options

Style Transfer Flow

1. Receive Reference Image

   • User provides an image file path

2. Analyze Reference

   • Use vision to understand the style
   • Extract key visual characteristics

3. Combine with New Subject

   • User specifies what they want in this style
   • Merge style analysis with new subject

4. Generate with Reference

   • Pass reference image to the script
   • Include style preservation directives in prompt

5. Iterate if Needed

   • Adjust based on feedback
   • Fine-tune style matching

Best Practices

DO

• Write prompts like a Creative Director describing intent
• Include specific technical details (lens types, lighting setups)
• Use concrete visual language, not abstract concepts
• Specify color palettes with descriptive terms
• Include composition guidance (rule of thirds, centered, etc.)
• Add quality markers (cinematic, photorealistic, studio quality)

DON'T

• Use "tag soup" (random comma-separated keywords)
• Include vague terms like "beautiful" or "amazing"
• Forget the purpose/use case context
• Ignore aspect ratio requirements
• Skip lighting descriptions

Prompt Templates by Category

App Icons

[Style] 3D app icon design
[Subject] [object] with [details]
[Composition] centered, rounded superellipse container
[Lighting] soft studio lighting with ambient occlusion
[Atmosphere] [color palette], clean modern aesthetic

Website Heroes

[Style] [photography/3D/illustration]
[Subject] [main visual element]
[Composition] wide 16:9, [focal point position]
[Lighting] [dramatic/soft/natural] with [color] accents
[Atmosphere] [mood], professional web aesthetic

Marketing Banners

[Style] commercial [photography/illustration]
[Subject] [product/concept] with [context]
[Composition] [aspect ratio], space for text overlay on [side]
[Lighting] [bright/moody], product-focused
[Atmosphere] [brand colors], persuasive visual impact

Social Media

[Style] [trending aesthetic]
[Subject] [engaging visual]
[Composition] square 1:1, scroll-stopping composition
[Lighting] [vibrant/soft]
[Atmosphere] [platform-appropriate mood], shareable quality

Iterative Refinement

When users request changes:

1. Identify the Adjustment

   • Color changes: "warmer", "cooler", "more vibrant"
   • Style changes: "more minimal", "more detailed"
   • Composition: "zoom in", "add more space"
   • Mood: "brighter", "more dramatic"

2. Modify the Prompt

   • Add specific adjustment language
   • Keep successful elements
   • Emphasize the change needed

3. Re-generate

   • Use updated prompt
   • Maintain same aspect ratio unless requested otherwise

Requirements

Before using this skill, ensure:

• Python 3.10+ is installed
• Required packages installed (see below)
• API key configured for your chosen provider

Install Dependencies

# For Google API
pip install google-genai Pillow

# For OpenRouter API
pip install openai Pillow

Provider Configuration

The plugin supports two API providers. Set via environment variable:

# Choose provider (add to ~/.zshrc or ~/.bashrc)
export NANO_BANANA_PROVIDER="google"    # or "openrouter"

Option 1: Google API (Recommended)

Best for image generation with full Nano Banana Pro features.

1. Go to https://aistudio.google.com
2. Sign in with your Google account
3. Click "Get API Key" -> "Create API Key"
4. Add to your shell profile:

   export GEMINI_API_KEY="your-key-here"
   export NANO_BANANA_PROVIDER="google"
   

5. Restart terminal or run source ~/.zshrc

Option 2: OpenRouter API

Alternative provider with unified API access.

1. Go to https://openrouter.ai
2. Sign in and navigate to Keys section
3. Create a new API key
4. Add to your shell profile:

   export OPENROUTER_API_KEY="your-key-here"
   export NANO_BANANA_PROVIDER="openrouter"
   

5. Restart terminal or run source ~/.zshrc

Provider Comparison