Nano Banana Pro
A drop-in Claude Code skill that generates custom images for your websites using Google's Gemini 3 Pro Image — straight from a natural-language prompt.
What it can make
Hover any image to see the JSON that made it
What you get
SKILL.md
Tells Claude when and how to invoke the generator — picks aspect ratios, paths, and writes great prompts.
generate_image.py
The CLI. Calls gemini-3-pro-image-preview, supports text-to-image plus reference-image editing.
install.sh
One-shot dependency installer for google-genai and pillow.
README.md
Human-facing setup guide with examples, flags, troubleshooting, and the Vertex AI option.
Install
-
Unzip into your Claude Code skills folder
# after downloading unzip nano-banana-pro.zip -d ~/.claude/skills/ -
Install the Python dependencies
bash ~/.claude/skills/nano-banana-pro/install.sh
-
Get an API key and export it
Grab one (free tier available) at aistudio.google.com/apikey, then add to your shell rc:
export GEMINI_API_KEY=your_key_here -
Restart Claude Code, then just ask
The skill auto-triggers on phrases like "generate a hero image", "make an OG card", "create an illustration of". Claude picks the prompt, params, and output path, runs the script, and wires the image into your code.
Use it directly
# text-to-image python3 ~/.claude/skills/nano-banana-pro/generate_image.py \ "A wide cinematic shot of a misty pine forest at sunrise" \ -o public/hero.png --aspect-ratio 16:9 --size 2K # edit / compose with a reference image python3 ~/.claude/skills/nano-banana-pro/generate_image.py \ "Place this logo centered on a deep navy background" \ -o public/og.png --aspect-ratio 1:1 --input assets/logo.png