ControlNet
Open-source neural network add-on for Stable Diffusion that gives precise control over image composition, pose, and structure.
About
ControlNet is a research project and open-source extension for Stable Diffusion that adds conditioning inputs alongside a text prompt to guide image generation with much greater precision. Instead of relying on words alone, you can supply a pose skeleton, edge map, depth map, segmentation mask, or other structural hints, and the model uses that information to control the layout and structure of the output image. This makes it practical for tasks like generating images with specific character poses, matching an architectural layout, or preserving the composition of a reference photo. ControlNet is free to use and runs locally, typically integrated into Stable Diffusion WebUI or ComfyUI.