Style transfer operations are the process for transferring a specific style of an image to the target image, while preserving the visual contents of the target image. Currently, the style transfer operations are usually implemented with machine learning techniques. Our goal is to finally achieve real-time style-transfer operations, as the special effect tools on the game engines, which is typically composed of various modules required to develop games. To efficiently implement the style transfer operations, as a special kind of machine-learning-related operations, we need to maintain efficient data management methods and also efficient handling of neural-network related operations. In this work, we designed a separate data processing method with Python modules, which are specialized for inferencing, to preserve the performance of the underlying game engine, as much as possible.