AI for Creative Visual Content Generation Editing and Understanding

Abstract

Generative AI has rapidly expanded the range of visual content that artists, designers, filmmakers, technical directors, and researchers can produce. Yet current generative workflows still face a major precision bottleneck: purely prompt-based systems often struggle to preserve spatial layouts, character identity, typography, scale relations, data values, and other constraints that are essential in professional creative production. This course introduces a structured approach to controllable AI creative production by combining multi-agent systems, computer graphics, and generative models. We present a shift from prompting to orchestrating. In this course, participants will learn how to decompose creative tasks into specialized agents, separate creative exploration from rigorous constraint generation, and connect generative models with graphics-based control signals such as depth maps, pose maps, edge maps, layouts, and 3D scene structures. The course is designed for the SIGGRAPH community broadly, with examples from filmmaking, digital illustration, photo post-production, graphic design, and data storytelling. The course emphasizes both conceptual understanding and practical workflow design. Through live demonstrations, cross-domain case studies, and step-by-step pipeline breakdowns, attendees will learn how to construct reproducible workflows that combine the visual richness of generative AI with the structural precision of graphics systems. The course also discusses how these workflows may expand creative industries by creating new roles such as AI pipeline designer, multi-agent workflow designer, AI art supervisor, and data-to-visual translator. Our goal is not to present AI as a replacement for artists, but to show how controllable AI systems can support more precise, collaborative, and extensible forms of creative production.