The image you get back is not a copy of your picture. A detector reads your picture first and draws a control map from it: outlines, a clean line drawing, how far away each part is, which way each surface faces, where the joints of a body are, flat colour fields, or the picture itself. The image the worker makes then follows that map.So choosing a control type is choosing what has to survive from your picture. Outlines keep the shapes and let the colours, the materials and the style go. Depth keeps the arrangement in space and lets the drawing go. Pose keeps the people and nothing else.The horde serves ControlNet on Stable Diffusion 1.5 models. Nine control types are served by any worker that offers ControlNet at all. The rest need a recently updated worker, so those requests reach fewer workers and can wait longer.How closely the image follows the map is a separate setting, the control strength, and its useful range is narrower than it looks: too low and the map is only a suggestion, too high and the image traces it.