banana

name	banana
description	Generate or edit images via Google Gemini (Nano Banana Ultimate). Text-to-image generation and multi-image editing with up to 14 input references. Triggers on: generate image, create image, edit image, make a picture, draw, render, visualize, banana, create visual.
allowed-tools	Bash(uv:*), Bash(ls:*), Read

Nano Banana Ultimate — Image Generation

Input: $ARGUMENTS (optional plain-language description of what to generate)

If $ARGUMENTS is provided, use it as the user's intent and skip to prompt building. If empty, ask a clarifying question.

Requirements

• uv installed
• GEMINI_API_KEY environment variable set (get one at https://aistudio.google.com/apikey)

Workflow

1. Understand — Parse the user's request. If the subject or intent is unclear, ask ONE clarifying question.
2. Build prompt — Construct a detailed Gemini prompt using the formula below.
3. Configure — Choose aspect ratio, resolution, and model. Use defaults unless the user specifies otherwise.
4. Generate — Run the script.
5. Deliver — Report the saved file path. Do NOT read the image file back. Offer to iterate: adjust style, composition, or prompt.

Prompt Formula

Build every prompt using this structure:

[Subject] + [Action/Pose] + [Environment/Location] + [Composition/Framing] + [Style/Medium]

Rules

• Positive framing: Describe what IS in the image, never what is absent. Say "empty street" not "no cars".
• Strong verb opener: Start with Generate, Create, Capture, Render, Design, Compose.
• Be specific: Include concrete details — materials, lighting direction, time of day, textures.
• Camera/lens (photorealistic): "shot on 85mm f/1.4 lens", "low angle", "aerial view".
• Lighting: "golden hour side light", "studio softbox", "dramatic chiaroscuro", "overcast diffused".
• Materiality: "brushed aluminum", "matte ceramic", "glossy lacquer", "translucent glass".
• Text in images: Enclose any text to render in quotes. Specify font style. Note: text rendering may be imperfect.

Generation

uv run "${CLAUDE_PLUGIN_ROOT}/scripts/banana.py" \
  --prompt "<detailed prompt>" \
  --output "<YYYY-MM-DD-HH-MM-SS-descriptive-name>.png" \
  --resolution <1K|2K|4K> \
  --aspect-ratio <ratio> \
  --model <model>

Editing (with input images)

When the user provides an existing image to modify:

uv run "${CLAUDE_PLUGIN_ROOT}/scripts/banana.py" \
  --prompt "<edit instruction: what to change AND what to keep>" \
  --output "<output-path>.png" \
  --input-image "/path/to/source.png"

For edits, describe both the change and what must stay identical:

"Change the background to a sunset beach. Keep the subject, their clothing, and pose exactly the same."

Defaults

Parameter	Default
Resolution	1K
Aspect ratio	1:1
Model	gemini-3-pro-image-preview

Models

Model	Best for
gemini-3-pro-image-preview	Final output, complex scenes, high quality (Nano Banana Pro)
gemini-3.1-flash-image-preview	Fast drafts, iterations, simple images (Nano Banana 2)

Aspect Ratios

Ratio	Use case
1:1	Social squares, logos, icons
3:2	Photography standard
2:3	Book covers, tall portraits
4:3	Landscape, presentations
3:4	Portrait photos, Pinterest
16:9	YouTube thumbnails, widescreen
9:16	Stories, Reels, TikTok
4:5	Instagram portrait
5:4	Large format print
21:9	Ultrawide banners

Output Naming

Always use timestamped filenames: YYYY-MM-DD-HH-MM-SS-descriptive-name.png

Save to the current working directory unless the user specifies a path.

References

See references/prompts.md for ready-to-use example prompts covering landscapes, abstract art, interiors, sci-fi, nature, fantasy, still life, and underwater scenes.