GenCAD

GenCAD is an AI model that generates parametric CAD programs and 3D solid models directly from images, addressing limitations of simpler representations like meshes or point clouds that lack the precision and editability needed for engineering tasks. The system combines transformer-based encoding, contrastive learning for multi-modal representation, latent diffusion models, and a decoder to produce complete, editable CAD command sequences rather than just 3D geometry. This approach enables more accurate automated design processes with the ability to generate multiple design variations from a single image input.