# gemini_image

> Generate images with Google Gemini (Nano Banana Pro) via CLI helper

- Author: Lloyd Vickery
- Repository: LloydVickeryASI/claude-code-skills-test
- Version: 20251222093431
- Stars: 0
- Forks: 0
- Last Updated: 2026-02-07
- Source: https://github.com/LloydVickeryASI/claude-code-skills-test
- Web: https://mule.run/skillshub/@@LloydVickeryASI/claude-code-skills-test~gemini_image:20251222093431

---

---
name: gemini_image
description: Generate images with Google Gemini (Nano Banana Pro) via CLI helper
---

# Gemini Image Skill

Quickly render images with Google Gemini using the Nano Banana Pro preview model.

## Dependencies
- `python3`
- `google-genai` (`pip install google-genai`)
- API key in `GOOGLE_GENAI_API_KEY` or `GEMINI_API_KEY`

## CLI Helper
`~/bin/gemini-image` (wrapper around `generate_image.py`)

```bash
# basic
gemini-image -p "A playful futuristic banana drone over a neon city"

# custom aspect ratio + output path
gemini-image -p "bioluminescent forest fox" -a 3:4 -o ~/tmp/fox.png

# alternate model
gemini-image -p "isometric data center" -m models/nano-banana-pro-preview
```

Flags:
- `-p/--prompt` (required): text prompt
- `-o/--output`: output file (default `~/tmp/gemini_image.png`)
- `-a/--aspect-ratio`: e.g., `1:1`, `3:4`, `16:9`
- `-m/--model`: default `models/nano-banana-pro-preview`
- `-k/--api-key`: override env var
- `--seed`: optional deterministic seed (int)
- `-r/--reference`: one or more reference images for style/composition guidance

```bash
# with reference images (e.g., logo for brand consistency)
gemini-image -p "Corporate banner with company branding" \
  -r ../marketing/assets/logos/ASI\ Logo\ White.png \
  -o ~/tmp/branded_banner.png
```

## Prompting Guide

For comprehensive prompting tips and examples, see the **[Nano Banana Prompting Guide](../marketing/nano-banana-prompting-guide.md)** which covers:

- Photorealistic scenes with photography terminology
- Stylized illustrations & stickers
- Accurate text rendering in images
- Product mockups & commercial photography
- Minimalist & negative space design
- Sequential art (comics/storyboards)
- Advanced techniques (multi-turn iteration, resolution control)

**Quick tip:** Describe scenes narratively rather than listing keywords - the model understands natural language better than tag-style prompts.

## Reference Images

Marketing assets are available at `../marketing/assets/` for use as reference images when generating brand-consistent content:

```
assets/
├── branding/      # Brand guidelines and style references
├── flyers/        # Existing flyer designs for style matching
├── linkedin-examples/  # LinkedIn post examples
└── logos/         # Company logos for inclusion
```

Nano Banana Pro supports up to 14 reference images to influence style and composition:
- Use 5-6 images for object/product consistency
- Use up to 5 images for character consistency
- Reference images help maintain brand colours, typography style, and visual identity

## Notes
- Saves PNG to the specified path; exits non-zero on failures.
- Model supports `response_modalities=['IMAGE']` with `image_config.aspect_ratio`.
- Keep prompts under ~500 tokens for fastest results.