What is this?
Nano Banana lets you create and edit images without ever leaving Claude Code. Just type what you want in plain English and the plugin handles the rest — picking the right model, enhancing your prompt, and saving the result.
/genimage a photorealistic cyberpunk street in Cairo, neon palm trees, 16:9
That's a real command. It works exactly like that. No config files, no boilerplate, no switching between tools.
Under the hood it uses Google's Gemini 3.1 Flash for fast generation and Gemini 3 Pro when you need high-resolution output. You don't need to think about which model to use — Nano Banana picks the right one automatically.
Part of the Ibrahim Plugins collection for Claude Code, alongside SupaConductor for multi-agent orchestration.
Features
- Text-to-image — describe anything, get an image back
- Image editing — "replace the sky with a sunset" and it just works, no masks needed
- Style transfer — apply the look of one image to another
- Multi-image composition — blend up to 14 reference images for consistent characters and scenes
- High-resolution output — go up to 4K for professional assets
- Search-grounded generation — images informed by real-time Google Search data
- Every aspect ratio — 1:1, 16:9, 9:16, 4:3, 21:9, and more
- Iterative editing — keep refining your image through conversation
All of this runs through a single /genimage command. The agent figures out which mode you need from context.
Installation
The quickest way:
claude plugin add Ibrahim-3d/nano-banana-claude-plugin
Or clone it yourself:
git clone https://github.com/Ibrahim-3d/nano-banana-claude-plugin.git ~/.claude/plugins/nano-banana
pip install google-genai python-dotenv Pillow
Restart Claude Code after installing to activate the plugin.
Get Started
You'll need a Gemini API key. It's free — grab one at Google AI Studio if you don't have one already.
Then inside Claude Code:
# Step 1 — save your API key (one-time setup)
/nano-banana:setup
# Step 2 — generate your first image
/genimage a golden retriever puppy playing in autumn leaves, soft bokeh, 4:3
That's it. Your image gets saved to the current directory. From here you can keep generating, edit what you've made, or try any of the examples below.
Usage Examples
Create images from text
/genimage a minimalist flat-design illustration of a coffee shop, pastel colors
/genimage product shot of a black ceramic mug on a marble surface, studio lighting, 1:1
/genimage aerial view of a sci-fi megacity at night, neon lights, cinematic, 21:9
Go high-res (2K / 4K)
Just mention the resolution — the plugin switches to the Pro model automatically:
/genimage a hyper-detailed fantasy castle on a cliff, waterfall, 4K, 16:9
Edit existing images
All editing is text-guided. Describe the change you want and Nano Banana figures out what to modify:
# Swap out a background
/genimage replace the plain background in photo.png with a bustling Tokyo street at night
# Remove something
/genimage remove the power lines from landscape.png, preserve everything else
# Transfer a style
/genimage apply the brushstroke style of starry-night.png to my-photo.jpg
Iterate on your work
Generate a base image, then keep refining it:
# Start with a scene
/genimage a cozy cafe interior, warm lighting, 4:3
# Then add to it
/genimage add a cat sleeping on the counter --images generated_image.png
You can chain as many edits as you like — each one builds on the last.
Ibrahim-3d
