Models and settings

Models

Each model is a different checkpoint with its own aesthetic. Some lean anime, some more realistic. There is no “best” one — pick by the look you want.

Every model has its own tuned settings: sampler, step count, guidance scale and prompt format. These are configured per model and applied automatically, so you do not need to know which is which.

Advanced settings

Available on paid plans:

Steps — how many refinement passes. More steps means more detail, up to a point, and a longer wait.
CFG / guidance — how strictly the model follows your prompt. Lower is more natural and creative; higher is more literal but can look forced.
Sampler and scheduler — the algorithm used.
Negative prompt — extra things to avoid, on top of the defaults.

Each model has a supported range, and values outside it are clamped to keep output coherent.

If you are unsure, leave these alone. The per-model defaults are tuned from each checkpoint’s own documentation.

How long it takes

Rendering typically takes 12–15 seconds. At busy times your job may queue behind others — paid plans are prioritised.