home / skills / idanbeck / claude-skills / nano-banana-pro

nano-banana-pro skill

safe

This skill helps you generate AI images and edit visuals using Gemini 3 Pro, delivering high quality results from prompts and reference images.

npx playbooks add skill idanbeck/claude-skills --skill nano-banana-pro

Review the files below or copy the command above to add this skill to your agents.

Files (2)

SKILL.md

2.5 KB

---
name: nano-banana-pro
description: Generate images using AI. Use when the user asks to create, generate, or make images, pictures, graphics, illustrations, visuals, or artwork. Also use for image editing with reference images.
allowed-tools: Bash, Read, Write
---

# Nano Banana Pro - AI Image Generation

Generate images using Google's Gemini 3 Pro Image model.

## Usage

Run the generation script:

```bash
python ~/.claude/skills/nano-banana-pro/generate_image.py "your prompt here" [options]
```

## Options

| Option | Description | Default |
|--------|-------------|---------|
| `--resolution` | Output size: 1K, 2K, or 4K | 2K |
| `--aspect` | Aspect ratio: 16:9, 1:1, 4:3, 9:16, 3:4 | 16:9 |
| `--output` | Output directory path | ./generated_images |
| `--reference` | Reference image(s) for style/editing (up to 14) | None |
| `--format` | Output format: png, jpeg, webp | png |

## Examples

### Basic generation

```bash
python ~/.claude/skills/nano-banana-pro/generate_image.py "a serene mountain landscape at sunset, photorealistic"
```

### Square image for social media

```bash
python ~/.claude/skills/nano-banana-pro/generate_image.py "abstract geometric pattern in blue and gold" --aspect 1:1 --resolution 2K
```

### High-res with custom output

```bash
python ~/.claude/skills/nano-banana-pro/generate_image.py "futuristic city skyline" --resolution 4K --output ~/Pictures/ai-generated
```

### Style transfer with reference image

```bash
python ~/.claude/skills/nano-banana-pro/generate_image.py "transform this into a watercolor painting" --reference input.jpg
```

### Image editing with reference

```bash
python ~/.claude/skills/nano-banana-pro/generate_image.py "add a rainbow in the sky" --reference landscape.png
```

## Output

Images are saved with timestamp filenames:
- Format: `{timestamp}_{sanitized_prompt}.{format}`
- Example: `20260106_143052_serene_mountain_landscape.png`

The script outputs the full path to the generated image.

## Requirements

- Python 3.10+
- `google-genai` library
- `GEMINI_API_KEY` environment variable

## Notes

- Gemini 3 Pro Image requires paid billing (no free tier)
- Generated images include SynthID watermarking
- Reference images enable style transfer and editing capabilities
- **No text by default**: The script automatically adds "no text/typography" instructions unless your prompt explicitly mentions text, typography, words, or labels. To include text, add words like "with text overlay" or "include typography" to your prompt.

Overview

This skill generates high-quality images using an advanced image model. It supports direct image creation and reference-based editing, with configurable resolution, aspect ratio, and output format. The skill is optimized for photorealistic, stylized, and illustrative results and saves outputs with timestamped filenames.

How this skill works

You provide a text prompt and optional reference images; the skill sends them to the image model and returns one or more generated files. Options let you choose resolution (1K, 2K, 4K), aspect ratio, output directory, file format, and up to 14 reference images for style transfer or edits. The script enforces a default "no text/typography" rule unless your prompt explicitly requests text, and it prints the full path to each saved image.

When to use it

Create original visuals for web, social, or print.
Generate concept art, scenery, or character studies from text prompts.
Perform style transfer or apply artistic edits using reference images.
Add elements to an existing photo (editing) by supplying a reference file.
Produce consistent outputs with controlled resolution and aspect ratio.

Best practices

Write clear, specific prompts describing subject, mood, lighting, and style.
Include explicit instructions for text or labels if you want typography included.
Use high-quality reference images for faithful style transfer or precise edits.
Select resolution and aspect ratio that match your final use (2K default for balance).
Keep reference image count reasonable; up to 14 supported but fewer often yields more coherent results.

Example use cases

Generate a photorealistic sunset mountain landscape for a website hero image.
Create social media-ready square artwork by setting aspect to 1:1 and resolution to 2K.
Produce a 4K futuristic city skyline for desktop wallpaper or concept presentation.
Convert a photo into a watercolor-style painting using a single reference image.
Edit a landscape to add a rainbow or remove elements by providing the original image as a reference.

FAQ

Do generated images include watermarks?

Yes. Generated images include SynthID watermarking as part of the model output.

How do I include text in the image?

Add explicit wording to your prompt such as "with text overlay" or "include typography"; otherwise the tool removes text by default.

What are the requirements to run the script?

You need Python 3.10+, the google-genai library, and a GEMINI_API_KEY environment variable with billing enabled for the model.