openrouter-image-generation

OpenRouter Image Generation

Use the image tools provided by the openrouter-genimage-kote plugin.

Image-to-image is the primary workflow.

Available tools

openrouter_edit_image

Use this tool when one or more existing images should influence the result.

Typical cases:

• edit an existing image;
• create a variation;
• redraw or restyle an image;
• change a background;
• recolor an asset;
• add or remove an object;
• preserve a character while changing the pose;
• preserve an object while changing its state;
• create another asset in the same visual style;
• combine subject, clothing, pose, palette, or style references;
• continue iterating on a previously generated image.

The tool requires at least one reference:

• a local file available inside the current Claude Code workspace roots;
• or a public HTTPS URL.

openrouter_generate_image

Use this tool only when the user requests a completely new image and no existing image needs to be preserved or used as a visual reference.

When to use the tools

Use an image tool when:

1. The user explicitly asks to generate, create, draw, render, edit, transform, redraw, or restyle an image.
2. The user requests new visual assets.
3. The user requests game assets such as:
   • character art;
   • sprites;
   • item icons;
   • ability icons;
   • textures;
   • backgrounds;
   • environment art;
   • concept art;
   • loading screens;
   • promotional images.
4. The user requests application or website assets such as:
   • UI illustrations;
   • empty-state illustrations;
   • thumbnails;
   • banners;
   • avatars;
   • hero images;
   • covers;
   • decorative graphics.
5. The user provides a usable image path or image URL and requests a visible modification.

The request may be written in any language. Follow the user's intent rather than exact English keywords.

When not to use the tools

Do not call an image tool when:

• the user only asks to discuss an idea;
• the user asks to analyze an image without changing it;
• the user asks to find an existing image online;
• the user asks only for factual information;
• the task should be solved with deterministic SVG, HTML, CSS, a chart, or a technical diagram;
• the user has not requested a new or modified visual result.