Real-Time Neural Style Transfer in Game Engines: A Comprehensive Review

Abstract

Neural Style Transfer (NST) has evolved into a transformative technique in computer vision and computer graphics, enabling the integration of artistic styles into digital media. In recent years, the application of NST in real-time scenarios—especially within game engines—has attracted significant academic and industrial interest. This paper provides a comprehensive review of the development of NST from its origins in 2D image transformation to its extension in video, 3D meshes, and real-time rendering. We further examine the role of modern game engines in supporting NST pipelines, summarize key advances in integrating stylization within interactive environments, and analyze both the benefits and limitations of such approaches. In addition, the paper highlights the existing research gaps, challenges, and possible future directions for advancing real-time NST in games.