Model

The trained network a picture is generated with.

Speed depends on the choice Kudos depend on the choice

Every picture the studio makes comes from one model, a network trained on images and their descriptions. Models differ in what they draw well and in the settings that suit them. Workers choose which models to load, so the list you can pick from is whatever the horde is running at the time.Each model descends from a baseline, the family it was trained from, such as SD 1.5, SDXL or Flux. The baseline decides which sizes and samplers suit the model and which LoRAs and embeddings work with it. It also changes how the kudos cost scales. The model list shows each model's baseline.A style is a saved bundle of a model, some settings and a prompt template your prompt is placed into. Community styles are shared on the horde and applied by the horde itself when the request arrives, so the settings a style sets take the place of the panel's. Your own saved copy of the panel is a preset, which stays private to this browser.

In the glossary