AI-Driven Workflow for Optimizing the Crowd Generation process in the VFX industry

The rapid advancement of Artificial Intelligence (AI) and Machine Learning (ML) has profoundly transformed the Visual Effects (VFX) industry, redefining traditional workflows and enabling the creation of hyperrealistic digital content with unprecedented efficiency. This thesis investigates the integration of AI-driven methodologies into the crowd generation process within VFX, focusing on a full-CG stadium crowd as a case study. Conducted in collaboration with EDI – Effetti Digitali Italiani, this research addresses key challenges in digital crowd animation and proposes an innovative AI-enhanced pipeline designed to optimize both realism and production efficiency. The study begins with a comprehensive review of state of the art AI applications in VFX, with a particular emphasis on 3D asset generation and animation. It then presents an in-depth analysis of the SideFX Houdini crowd generation process, identifying areas where AI can streamline workflows and enhance the overall quality. A core component of this research is the exploration of AI-powered tools—including Meshcapade, MoveAI, Rokoko Vision, and GenMM—to evaluate their effectiveness in animating digital characters. Special attention is given to AI Markerless Motion Capture and the post-processing workflow required to enhance the results, along with the crucial retargeting step necessary for transferring the animation to the crowd agents' skeletons. The proposed workflow seamlessly integrates AI-driven animation techniques with industry-standard software SideFX Houdini, demonstrating a substantial reduction in production time while maintaining high visual fidelity. Through a comparative evaluation of both quantitative metrics and qualitative assessments, this study highlights the advantages of AI-assisted crowd generation, particularly in streamlining animation workflows and enhancing the believability of digital crowds. By contributing to the evolving landscape of AI-assisted VFX production, this research underscores the transformative potential of AI in the VFX industry. It emphasizes the importance of balancing automation with artistic control, ensuring that AI serves as an augmentative tool rather than a replacement for human creativity.