Nano Banana Image Generation Skill

Generate and edit images using Google's Nano Banana (Gemini 2.5 Flash Image) and Nano Banana Pro (Gemini 3 Pro Image) APIs.

Model Selection

Model	ID	Best For	Resolution	Cost
Nano Banana	gemini-2.5-flash-image	Fast generation, iteration, basic edits	Up to 1024px	~$0.039/image
Nano Banana Pro	gemini-3-pro-image-preview	Professional assets, text rendering, complex compositions	Up to 4K	Higher cost

Selection Guide:

• Use Nano Banana for: rapid prototyping, simple edits, high-volume generation
• Use Nano Banana Pro for: text in images, 4K output, up to 14 reference images, Google Search grounding

Core Capabilities

1. Text-to-Image: Generate images from text descriptions
2. Image Editing: Add, remove, modify elements in existing images
3. Multi-Image Composition: Blend up to 14 images (Pro only)
4. Character Consistency: Maintain same character across multiple generations
5. Text Rendering: Generate legible text in images (Pro excels here)
6. Style Transfer: Apply artistic styles to images
7. Iterative Refinement: Conversational multi-turn editing

Quick Start

Generate an Image

python scripts/generate_image.py "A cozy coffee shop interior with warm lighting" --output coffee_shop.png

Edit an Image

python scripts/edit_image.py input.jpg "Add a cat sitting on the chair" --output output.png

Prompting Best Practices

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

The model understands natural language narratives better than comma-separated tags.

Prompt Structure (for best results)

Include these elements in your prompts:

1. Subject: Who/what is in the image (be specific)
2. Action: What is happening
3. Environment/Location: Setting and context
4. Lighting: Natural, studio, golden hour, etc.
5. Style: Photorealistic, illustration, watercolor, etc.
6. Composition: Camera angle, framing, perspective
7. Mood/Atmosphere: Emotional tone

Example - Good vs Bad Prompts

Bad: cat, hat, wizard, cute

Good: A fluffy ginger cat wearing a tiny knitted wizard hat, sitting on a wooden floor in a cozy living room. Soft natural light streams through a nearby window, creating a warm, magical atmosphere. Photorealistic, shot with an 85mm portrait lens.

For comprehensive prompting strategies, see: references/prompting-guide.md

API Configuration

Required Environment Variable

export GEMINI_API_KEY="your-api-key-here"

Get your API key from: https://aistudio.google.com/apikey

Response Modalities

Always set responseModalities: ["TEXT", "IMAGE"] to receive generated images.

Image Configuration Options

image_config = {
    "aspect_ratio": "16:9",  # Options: 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9
    "image_size": "2K"       # Options: 1K, 2K, 4K (Pro only for 4K)
}

For complete API reference, see: references/api-reference.md

Scripts Reference

Script	Purpose
scripts/generate_image.py	Text-to-image generation
scripts/edit_image.py	Edit existing images with text prompts
scripts/multi_image_compose.py	Compose multiple images (Pro only)

Important Notes

• All generated images include invisible SynthID watermarks
• Pro model uses "thinking" mode for complex prompts (enabled by default)
• For multi-turn editing, maintain conversation history
• Supported input formats: JPEG, PNG, WebP (up to 5MB)
• For best performance use languages: EN, es-MX, ja-JP, zh-CN, hi-IN