name: nano-banana-image-generator
description: Use this agent when an article has been completed by the content-writer agent AND reviewed by the article-credibility-reviewer agent, and a visual header image is needed for the post. This agent should be triggered proactively after both content creation and credibility review are complete.

Nano Banana Image Generator

Overview

This skill provides comprehensive image generation and editing capabilities using Google's Nano Banana (Gemini 2.5 Flash Image) and Nano Banana Pro (Gemini 3 Pro Image) models. Generate complex UI mockups, hero images, illustrations, and any visual assets needed for articles, applications, or presentations.

Prerequisites

Before using this skill, ensure:
1. The GEMINI_API_KEY or GOOGLE_API_KEY environment variable is set
2. Required packages are installed (auto-installs on first run, or run setup manually)

Setup

The script automatically installs dependencies on first run using the best available installer:
- uv (preferred - 10-100x faster than pip)
- pip3 (fallback)
- pip (fallback)

To manually install or force reinstall:

# Auto-install using best available installer
python scripts/nano_banana.py setup

# Force reinstall
python scripts/nano_banana.py setup --force

# Or manually with uv (recommended)
uv pip install google-genai Pillow

# Or with pip
pip install google-genai Pillow

Quick Start

Generate Article Header Image

To generate a header image for a completed article:

python scripts/nano_banana.py generate "A professional hero image for an article about [topic], featuring [visual elements], [style] style" --model pro --aspect-ratio 16:9 --output header.png

Generate UI Mockup

python scripts/nano_banana.py ui "Dashboard showing user analytics with charts and metrics" --type web --style modern --output dashboard.png

Edit Existing Image

python scripts/nano_banana.py edit input.png "Make the colors more vibrant and add a professional gradient overlay" --output enhanced.png

Core Capabilities

1. Text-to-Image Generation

Generate images from text descriptions using either:
- Flash model: Fast generation (~3-5 seconds), good for iteration
- Pro model: Higher quality (~8-15 seconds), better for final output

from scripts.nano_banana import generate_image

# Basic generation
paths = generate_image(
    prompt="A futuristic city skyline at sunset with flying cars",
    model="flash",  # or "pro" for higher quality
    aspect_ratio="16:9",
    output_path="city.png"
)

2. UI Mockup Generation

Specialized function for generating realistic interface mockups:

from scripts.nano_banana import generate_ui_mockup

path = generate_ui_mockup(
    description="A social media analytics dashboard with follower growth charts, engagement metrics, and recent post performance",
    ui_type="dashboard",  # web, mobile, desktop, tablet, dashboard
    style="modern",  # modern, minimal, corporate, playful, dark, glassmorphism
    model="pro",
    output_path="social_dashboard.png"
)

Available UI types:
- web: Website interface (16:9)
- mobile: Mobile app interface (9:16)
- desktop: Desktop application (16:9)
- tablet: Tablet app interface (4:3)
- dashboard: Analytics dashboard (16:9)

3. Image Editing

Edit existing images with natural language instructions:

from scripts.nano_banana import edit_image

path = edit_image(
    image_path="original.png",
    prompt="Remove the background and replace with a gradient from blue to purple",
    model="flash",
    output_path="edited.png"
)

4. Multi-Image Blending

Combine multiple images into a cohesive composition:

from scripts.nano_banana import blend_images

path = blend_images(
    image_paths=["photo1.png", "photo2.png", "photo3.png"],
    prompt="Create a seamless collage combining elements from all images",
    model="flash",
    output_path="collage.png"
)

5. Style Transfer

Apply artistic styles to images:

from scripts.nano_banana import style_transfer

# Using text description
path = style_transfer(
    content_image_path="photo.png",
    style_reference="watercolor painting with soft edges and flowing colors",
    output_path="watercolor.png"
)

# Using reference image
path = style_transfer(
    content_image_path="photo.png",
    style_reference="style_reference.png",
    output_path="styled.png"
)

Article Header Image Workflow

When generating header images for articles:

1. Analyze article content: Identify key themes, concepts, and visual metaphors
2. Choose appropriate style: Match the article tone (professional, creative, technical)
3. Generate with Pro model: Use --model pro for publication-quality images
4. Use 16:9 aspect ratio: Standard for blog headers and social sharing

Example Prompts for Article Headers

Technical Article:

A professional hero image for a technical article about cloud architecture,
featuring abstract cloud formations with interconnected nodes and data streams,
clean modern style with blue and white color scheme, soft lighting

Business Article:

A sophisticated header image for a business strategy article,
featuring abstract geometric shapes representing growth and progress,
corporate professional style with navy and gold accents, gradient background

Creative Article:

An artistic hero image for a creative writing article,
featuring flowing ink strokes transforming into words and ideas,
watercolor style with vibrant colors, organic flowing composition

Supported Aspect Ratios

Model Selection Guide

Command Line Reference

# Setup dependencies (uses uv if available, pip as fallback)
python scripts/nano_banana.py setup
python scripts/nano_banana.py setup --force  # Force reinstall

# Generate image
python scripts/nano_banana.py generate "prompt" --model flash --aspect-ratio 16:9 --output image.png --count 4

# Edit image
python scripts/nano_banana.py edit input.png "editing instructions" --output edited.png

# Generate UI mockup
python scripts/nano_banana.py ui "description" --type mobile --style dark --output app.png

# Blend images
python scripts/nano_banana.py blend img1.png img2.png --prompt "blend instructions" --output merged.png

# Style transfer
python scripts/nano_banana.py style content.png --style "Van Gogh" --output styled.png

References

For detailed API documentation and advanced prompting techniques:
- references/api_reference.md: Complete API documentation
- references/ui_prompting_guide.md: UI mockup generation best practices

Troubleshooting

API Key Not Found

Ensure GEMINI_API_KEY or GOOGLE_API_KEY is set:

export GEMINI_API_KEY="your-key-here"

Rate Limiting

If encountering rate limits, wait 60 seconds between requests or use the Flash model for faster quota recovery.

Safety Filters

If content is blocked, rephrase the prompt to avoid triggering safety filters. Avoid requests for real people, violence, or copyrighted characters.