Generate and Edit Images with the Nano Banana Pro Skill

AI image generation has evolved rapidly, but integrating it into agent workflows often requires juggling APIs, SDKs, and complex prompts. The Nano Banana Pro skill brings Google's powerful Gemini 3 Pro Image model directly into your Clawdbot toolkit—enabling both text-to-image generation and image editing with simple commands.

Who Needs This?

If you've ever wanted your AI assistant to:

Generate custom images from descriptions on the fly
Edit existing images (add objects, change styles, adjust colors)
Iterate quickly on visual concepts before producing high-resolution finals
Automate visual content creation workflows

...then Nano Banana Pro is exactly what you need. It's particularly powerful for content creators, designers prototyping ideas, and developers building visual tools.

Installation

Install the skill with a single command:

clawdhub install nano-banana-pro

Or if using npx:

npx clawdhub@latest install nano-banana-pro

Prerequisites

uv (Python package runner) - the skill uses uv run to execute Python scripts
GEMINI_API_KEY - get one from Google AI Studio

Set your API key as an environment variable:

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

Usage Examples

Generate a New Image

The basic syntax for image generation:

uv run ~/.codex/skills/nano-banana-pro/scripts/generate_image.py \
  --prompt "minimalist logo with abstract mountain, geometric shapes, blue gradient" \
  --filename "logo-final.png" \
  --resolution 4K

This creates a 2048px image saved to your current directory.

Edit an Existing Image

Want to modify a photo? Use the --input-image flag:

uv run ~/.codex/skills/nano-banana-pro/scripts/generate_image.py \
  --prompt "minimalist logo with abstract mountain, geometric shapes, blue gradient" \
  --filename "logo-final.png" \
  --resolution 4K

Common editing tasks include:

Adding or removing elements
Changing artistic styles
Adjusting colors and lighting
Blurring backgrounds
Converting to illustration/cartoon styles

Resolution Options

Choose from three resolution tiers:

1K (~1024px) — Quick drafts, iteration
2K (~2048px) — Most use cases
4K (~4096px) — Final high-quality output

Pro tip: Always use 1K for drafts. Once your prompt is dialed in, regenerate at 4K.

The Draft → Iterate → Final Workflow

This is the secret to efficient image generation without burning time or API credits:

1. Draft (1K) — Get quick feedback on your concept:

uv run ~/.codex/skills/nano-banana-pro/scripts/generate_image.py \
  --prompt "minimalist logo with abstract mountain, geometric shapes, blue gradient" \
  --filename "logo-final.png" \
  --resolution 4K

2. Iterate — Refine the prompt based on results:

uv run ~/.codex/skills/nano-banana-pro/scripts/generate_image.py \
  --prompt "minimalist logo with abstract mountain, geometric shapes, blue gradient" \
  --filename "logo-final.png" \
  --resolution 4K

3. Final (4K) — Lock in your prompt, generate production quality:

uv run ~/.codex/skills/nano-banana-pro/scripts/generate_image.py \
  --prompt "minimalist logo with abstract mountain, geometric shapes, blue gradient" \
  --filename "logo-final.png" \
  --resolution 4K

Prompt Tips

For Generation

Pass your image description directly. The API handles creative interpretation well. Only rework prompts if results are clearly off.

For Editing

Be specific about what to change AND what to preserve:

"Change ONLY: add a rainbow in the sky. Keep identical: subject, composition, lighting, color palette, and overall style."

This prevents the model from unexpectedly altering other parts of your image.

Troubleshooting

"Error: No API key provided" — Set GEMINI_API_KEY or pass --api-key YOUR_KEY

"Error loading input image" — Check that --input-image points to a valid, readable image file

403 or quota errors — Your API key may lack access or have exceeded limits. Try a different key or check your Google AI Studio dashboard.

Conclusion

Nano Banana Pro transforms your agent into a visual content creator. The draft-iterate-final workflow keeps costs down while ensuring quality, and the editing capabilities mean you can refine images without starting from scratch.

Whether you're generating marketing assets, prototyping UI mockups, or just having fun with AI art—this skill delivers Google's best image model in a clean, scriptable package.

Links: