Adding nano-banana 3 Support to My CLI Wrapper
Twenty-four hours after Google dropped nano-banana 3, I shipped support for it. New model, new resolutions (4K!), new features. This is what building with AI is like at this point in November 2025.
My nano-banana CLI tool wraps Google's Gemini image generation API. When they released gemini-3-pro-image-preview, I added a `--model` flag to let users pick between version 2 and 3. Simple enough.
But model 3 came with new capabilities: 4K resolution options and multi-reference image support. So I added `--resolution 4K` and `--reference` flags. The tool now calculates costs for both model versions since pricing differs.
While testing, I noticed a bug I'd been living with: relative paths weren't saving to the `output/` directory. Fixed that too.
Using Claude Code the whole update took about 20 minutes to implement and test. Building with AI tools—the APIs move fast, but adding support is straightforward in well-structured and simple code.
I wasn't building something new. I was keeping pace with Google. The tools evolve so quickly that "maintenance" means shipping new features very quickly.
It was a lot of fun to use my AI enabled workflow to implement this change so quickly and be able to see results!
Below is an example of how I use my nano-banana-experiments CLI tool and a family photo that I edited using the tool.
CLI tool usage example
Original Photo
Edited image using the nano-banana-experiments CLI tool with the following CLI command “nano-banana generate "change the clothing of the people and dog to have them dressed in traditional Japanese clothes" -i ~/Projects/mosaic-mesh-ai-blog/assets/gottschalk-family-original.jpg -o gottschalk-family-nano-banana-3-edited.jpg --model 3 --resolution 1K”
What problems do you see in the AI edited image? Can you tell that it was AI edited? These are important question to consider. The goal of this CLI tool is to make interaction with the Gemini model simple from my local machine but it doesn’t attempt to make ethical or moral choices about what is appropriate use of such a tool.
nano-banana-experiments - View on GitHub
*This post is part of my daily AI journey blog at [Mosaic Mesh AI](https://www.mosaicmeshai.com/blog). Building in public, learning in public, sharing the messy middle of AI development.*