Dynamic sparkle rendering: photorealistic real-time visualization of glittering surfaces in interactive scenarios

Abstract The need for photorealistic rendering of glittery materials, motivated by applications in gaming, film, virtual reality, and particularly automotive visualization, requires real-time techniques to depict millions of minute flakes in paint coatings. Existing approaches often rely on a single grid for glint mapping, leading to repetitive patterns, aliasing, and noticeable popping during camera movement. Addressing these issues is essential for achieving convincing sparkle under varying lighting and viewpoints. We propose a new real-time method that integrates three procedurally generated microflake grids, each with multiple resolution levels, to ensure smooth transitions between levels of detail and eliminate tiling artifacts. Adaptive scaling dynamically adjusts sparkle size and brightness based on camera distance, while per-flake light-color interactions enhance reflection realism. Our custom shader framework, paired with an artist-friendly material editor, enables real-time adjustments to flake density, size, falloff, and color. Compared to single-grid or fixed-scale methods, our multi-grid approach delivers dense, non-repetitive sparkles with consistent quality across distances, albeit with higher GPU demands in complex scenes. The controllable and diverse output of our system also makes it well-suited for generating synthetic training data for AI models focused on material recognition, lighting estimation, or view-dependent appearance. Future work will explore deferred rendering, 3D noise textures, optimized shader branching, and temporal filtering to enhance performance and visual quality.