AI image generation CLI built in Go, powered by Gemini image models. Rewritten from nano-banana-2-skill.

Key differences from the original:

• Restructured codebase with full --help output
• Stdin pipe support (echo "prompt" | nano-banana)
• Human-readable logging with [nano-banana] prefix
• --costs --json includes per-model breakdown
• Pure Go green screen background removal (no external tools needed)
• Unit tests
• GitHub Actions CI / Release workflows

Requires: Go 1.25+

go mod tidy
go build -o ~/.local/bin/nano-banana ./cmd/nano-banana

Make sure ~/.local/bin is in your PATH.

Gemini API key is resolved in this order:

1. --api-key flag
2. GEMINI_API_KEY environment variable
3. .env in current directory
4. .env next to the executable's parent directory
5. ~/.nano-banana/.env

# Recommended: environment variable
export GEMINI_API_KEY=your_key_here

# Or use a dotenv file
mkdir -p ~/.nano-banana
echo "GEMINI_API_KEY=your_key_here" > ~/.nano-banana/.env

Get an API key: https://aistudio.google.com/apikey

nano-banana "minimal dashboard UI with dark theme"
nano-banana "luxury product mockup" -o product -s 2K
nano-banana "cinematic scene" -a 16:9 -s 4K
nano-banana "change to white background" -r input.png -o output
nano-banana "robot mascot" -t -o mascot

# Read prompt from stdin
echo "a cat in a spacesuit" | nano-banana
pbpaste | nano-banana -s 2K -o result

-t automatically adds green screen instructions to the prompt, then removes the background using a built-in pure Go implementation of colorkey + despill + trim. No external tools required.

nano-banana "robot mascot" -t -o mascot

Every generation is logged to ~/.nano-banana/costs.json:

nano-banana --costs
nano-banana --costs --json

make test     # Run tests
make build    # Build binary
make lint     # vet + format check
make ci       # Full CI pipeline