ControlNet Conditioning
Introduced by Lvmin Zhang and Maneesh Agrawala in 2023, ControlNet adds a trainable branch that steers a diffusion model with structure maps such as edges, depth or pose, so layout stays fixed while style changes.
Spot it
- 01Preserved outlines and pose
- 02Fully restyled surfaces
- 03Structure-locked redraw
Why it happens
A copy of the model's encoder is trained to read a control image, such as an edge map, and its signals are added into the main network so generation follows that structure.
Recreate it in Pixio
Image edit prompt“Redraw this image using its edge outlines and depth as strict structural guidance: every contour, pose and proportion preserved exactly, while surfaces, lighting and materials are fully restyled as a polished painted illustration with rich color. Keep the original subject and composition.”
Open PixioUpload your image and pick an image-edit model.