By name: google-gemini-image type: skill version: '2.1' description: Generate, edit, and refine images with Google Gemini. Load when user says 'generate image', 'edit image', 'refine image', 'text to image', 'gemini image', 'modify image'. category: integrations tags:
- gemini
- images platform: Google Gemini updated: '2026-03-23' visibility: public
Google Gemini Image
Unified skill for text-to-image generation, image editing, and iterative refinement.
Replaces: gemini-generate-image, gemini-edit-image, gemini-refine-image.
Usage
# Generate from text prompt
uv run python scripts/gemini_image.py generate "a cat in space"
uv run python scripts/gemini_image.py generate "sunset over mountains" --aspect 16:9 --size 2K
uv run python scripts/gemini_image.py generate "abstract art" --output my_art.png
# Edit an existing image
uv run python scripts/gemini_image.py edit photo.png "make the sky blue"
uv run python scripts/gemini_image.py edit scene.jpg "add clouds" --aspect 4:3 --size 1K --output result.png
# Refine the last generated/edited image
uv run python scripts/gemini_image.py refine "add more stars"
uv run python scripts/gemini_image.py refine "make it brighter" --image specific.png
Model
All actions use gemini-3.1-flash-image-preview (Nano Banana 2). Inputs are text-only, or image + text for edit/refine.
Legacy gemini-2.0-flash-exp* models are deprecated; Google scheduled shutdown June 1, 2026 — this skill targets 3.1 only.
Aspect ratio and size
--aspect:1:1,1:4,1:8,2:3,3:2,3:4,4:1,4:3,4:5,5:4,8:1,9:16,16:9,21:9(default1:1).--size:512,1K,2K,4K(default1K). Passed to the API asimage_config.image_size.
Requirements
GEMINI_API_KEYin.env(get from https://aistudio.google.com/app/apikey)pip install google-genai Pillow
Output
Images save as PNG to 04-workspace/generated-images/ (or --output path).
Filenames auto-generated with timestamp: generated_20260323_143000.png, edited_..., refined_....
Shared path helpers live in scripts/gemini_client.py.