Glossary term
Glossary term
Multimodal AI
Adapter that adds conditional spatial control to diffusion models using edge maps, depth maps, poses, or segmentation masks.
Adobe Photoshop's Generate Image feature uses ControlNet-style conditioning to generate content that matches the structural layout of the canvas - filling masked regions while respecting surrounding composition.
Interior design startups use ControlNet with depth maps from iPhone LiDAR scans to generate redesigned room images that preserve the existing room geometry while changing furniture, colours, and materials.
Pose-conditioned ControlNet is used by fashion e-commerce companies to generate model images in specific poses from a garment image and a pose skeleton - reducing commercial photo-shoot costs by 70%.