ControlNet is an add-on architecture that lets you constrain an AI image generation to a specific pose, depth map, edge map or composition. Instead of relying on the prompt alone, you feed the model a second input — a stick-figure pose, a Canny edge outline, a depth map — and the output respects that structure. ControlNet is the technical foundation of most professional AI image workflows because it lets a designer hold composition constant while iterating on style. Originally built for Stable Diffusion; Flux 2 has a compatible variant called Flux Controls.