Image Generation

Unlike text generation, there's no standard API for image generation across providers - each requires its own custom implementation. Despite the additional effort required, there's now seamless image generation support through both the UI and CLI with built-in integrations for:

To begin select an image generation model from the Model Selector that supports image generation:

When an image generation model is selected, the chat prompt will the option to specify which aspect ratio to use for the generated images:

Command-Line Usage

Generate images using the --out image modifier:

llms --out image "cat in a hat"

Which uses the out:image chat template in llms.json for its image generation request. Before returning, any assets are saved to cache and their local path and HTTP URL returned, e.g:

Output:

Here is a cat in a hat for you! 

Saved files:
/home/mythz/.llms/cache/c9/c9b2fd2a1d95708251...5d3f467a.png
http://localhost:8000/~cache/c9/c9b2fd2a1d95708251...5d3f467a.png

Specify a Model

Use any model that supports image generation by specifying its ID or name:

llms -m "gemini-2.5-flash-image" --out image "cat in a hat"
llms -m "Gemini 2.5 Flash Image" --out image "cat in a hat"