Nano Banana Skill

Nano Banana Skill

A Claude Code Skill for AI-powered image generation and editing using Google Gemini's Nano Banana models.

When to Use This Skill

Activate this skill when users request:

• Image generation from text descriptions
• AI image editing or modification
• Creating visual content with specific styles
• Generating images with custom aspect ratios

Trigger phrases:

• "Generate an image of..."
• "Create a picture of..."
• "Edit this image to..."
• "Make an image with..."
• "生成图片"
• "创建图片"
• "编辑图片"
• "AI 绘图"

Prerequisites

1. API Key: Set environment variable NANO_BANANA_API_KEY

   export NANO_BANANA_API_KEY="your-api-key"
   

2. Required tools: curl, jq, base64

Available Models

Quick Start

Text to Image Generation

# Basic generation
bash ~/.claude/skills/nano-banana-skill/scripts/generate.sh \
  -p "A cute cat wearing a space helmet on Mars" \
  -o ./output

# With aspect ratio
bash ~/.claude/skills/nano-banana-skill/scripts/generate.sh \
  -p "A panoramic mountain landscape at sunset" \
  -o ./output \
  -r 16:9

# Using Pro model with 4K output
bash ~/.claude/skills/nano-banana-skill/scripts/generate.sh \
  -p "A detailed portrait of a robot philosopher" \
  -o ./output \
  -m pro \
  -s 4K

Image Editing

# Edit an existing image
bash ~/.claude/skills/nano-banana-skill/scripts/generate.sh \
  -p "Add a rainbow in the sky" \
  -i ./input/photo.jpg \
  -o ./output

# Style transfer
bash ~/.claude/skills/nano-banana-skill/scripts/generate.sh \
  -p "Transform this photo into a watercolor painting" \
  -i ./input/photo.png \
  -o ./output \
  -m pro

Parameters

Aspect Ratios

Flash model (14 ratios): 1:1, 1:4, 1:8, 2:3, 3:2, 3:4, 4:1, 4:3, 4:5, 5:4, 8:1, 9:16, 16:9, 21:9

Pro model (10 ratios): 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9

Output Image Sizes

• 512 - 512px (Flash only)
• 1K - 1024px
• 2K - 2048px
• 4K - 4096px (highest resolution)

Examples

1. Generate a Logo

bash ~/.claude/skills/nano-banana-skill/scripts/generate.sh \
  -p "A minimalist logo for a coffee shop called 'Bean Dreams', modern flat design, warm colors" \
  -o ./logos \
  -r 1:1 \
  -m pro \
  -s 2K

2. Create Social Media Content

# Instagram post (square)
bash ~/.claude/skills/nano-banana-skill/scripts/generate.sh \
  -p "Aesthetic flat lay of coffee and pastries, soft morning light" \
  -o ./social \
  -r 1:1

# Instagram story (portrait)
bash ~/.claude/skills/nano-banana-skill/scripts/generate.sh \
  -p "Motivational quote background, gradient sunset colors" \
  -o ./social \
  -r 9:16

3. Edit Product Photos

bash ~/.claude/skills/nano-banana-skill/scripts/generate.sh \
  -p "Remove the background and add a professional studio backdrop" \
  -i ./product.jpg \
  -o ./edited \
  -m pro

4. Concept Art

bash ~/.claude/skills/nano-banana-skill/scripts/generate.sh \
  -p "Futuristic cyberpunk city street at night, neon lights reflecting on wet pavement, detailed concept art" \
  -o ./concept \
  -r 16:9 \
  -m pro \
  -s 4K

API Endpoints

• Primary: https://undyapi.com
• Backup 1: https://vip.undyingapi.com
• Backup 2: https://vip.undyingapi.site

For detailed API documentation, see REFERENCE.md.

Troubleshooting

API Key Not Found

Error: NANO_BANANA_API_KEY environment variable not set

Solution: Export your API key: export NANO_BANANA_API_KEY="your-key"

Rate Limited

Error: 429 Too Many Requests

Solution: Wait a moment and retry, or use the backup endpoints

Invalid Image Format

Error: Unsupported image format

Solution: Supported formats are JPEG, PNG, GIF, WebP

Best Practices

1. Be specific in prompts: Include style, mood, lighting, and composition details
2. Use Pro model for text rendering: Flash model may not render text accurately
3. Choose appropriate aspect ratios: Flash supports extreme ratios (1:8, 8:1), Pro does not
4. Use 512 for quick iterations: Flash-only, fastest generation

Output

Generated images are saved to the specified output directory with timestamped filenames:

• nano_banana_20260109_143052.jpg (extension matches actual format from API)

The script also outputs:

• The full path to the generated image
• Any text response from the model (if applicable)