ControlNet is a neural-network technique that adds spatial control to text-to-image diffusion models. Instead of only following a text prompt, the model also follows a “control image” such as edges, straight lines, a depth map or a segmentation map. It was introduced by Lvmin Zhang, Anyi Rao and Maneesh Agrawala in 2023.
Why it matters for interiors
A plain image model asked to “redesign this living room” tends to redraw everything: windows move, the ceiling rises, the camera shifts. With ControlNet, the tool first extracts the room’s lines or depth from your photo and makes the new image respect them. That is how an AI redesign can change the sofa and the wall colour while keeping the window exactly where it is.
Common control types used for rooms:
- Line detection (for example M-LSD), which captures straight architectural lines.
- Depth maps, which keep what is near and far.
- Segmentation, which labels wall, floor, ceiling, window and furniture.
In practice
ControlNet is one of several ways to keep structure; tools combine it with img2img and inpainting. In Tigmi, Room Lock does this job for you.