Image Generation Configs¶

Image generation uses both an image service config and an LLM chat config:

The image config sends the final prompt to the image backend, currently ComfyUI.
The chat config builds image prompts from world state, character state, or narration.

Use the Images tab in Configurations to create reusable ComfyUI image model configs.

ComfyUI image fields¶

Field	Description
`model`	Optional checkpoint/model name.
`vae`	Optional VAE model name.
`clip`	Optional CLIP model name.
`image_width` / `image_height`	Optional requested image dimensions.
`seed`	Optional deterministic seed.
`steps`	Optional generation step count.
`cfg`	Optional classifier-free guidance value.

Assignments¶

Assign image configs to world or simulation image components:

Component	Purpose
`character_image_generator`	Character state/cover images.
`character_portrait_image_generator`	Character portrait images.
`location_image_generator`	Location images.
`item_image_generator`	Item images.
`scene_image_generator`	Turn or scene illustrations.

For automatic turn images, make sure both scene_image_generator and turn_image_trigger have suitable chat model assignments, and that scene_image_generator also has an image model assignment.

Per-simulation behavior¶

Per-simulation image generation behavior is controlled by ImageGenerationConfig:

Field	Description
`mode`	`manual`, `auto`, or `always`. Manual only generates on request. Auto uses a significance check and fallback interval. Always generates every turn.
`fallback_turns`	In `auto` mode, force an image after this many turns without one.