Browse Teach Curate

Conditional Image Generation with ControlNet

Adding spatial control to a pretrained text-to-image diffusion model by conditioning generation on an auxiliary structural input such as an edge map, pose, or segmentation mask, so the output follows the provided layout while the base model supplies appearance.

jeremy Video

ControlNet Architecture and PyTorch Implementation in Diffusion Models

ControlNet architecture introduces a mechanism to inject spatial conditional information into pre-trained diffusion models without degrading their generative capabilities learned from massive dataset…