This project aims to enhance the visual diversity and realism of crowds in real-time virtual environments by combining AI-driven texture generation for clothing with procedural variation techniques applied to a limited set of template MetaHumans in Unreal Engine. By leveraging state-of-the-art AI models to generate diverse, high-quality clothing textures dynamically, the project will create varied and authentic appearances for crowd characters without the need for extensive manual asset creation. The student will develop a pipeline that integrates AI texture synthesis with MetaHuman customization and crowd simulation, evaluating the system’s effectiveness in producing visually convincing and diverse crowds in real-time. This approach has applications in gaming, virtual production, and large-scale simulations where realistic crowd representation is essential but asset budgets are limited.