ControlNet
An add-on for diffusion image models that conditions generation on an input such as a pose, depth map, or edge outline, giving precise control over composition.
A plain text prompt gives a diffusion model a lot of freedom over composition, which is not always what you want. ControlNet lets you constrain generation with an extra input, for example a stick-figure pose, a depth map, a scribble, or the edges of a reference image, so the output follows that structure while the prompt fills in the detail.
It is one of the reasons the SDXL ecosystem is so deep: there are ControlNet models for many kinds of conditioning, and tools like ComfyUI and AUTOMATIC1111 support them directly. It builds on diffusion.