Constructing action scenes for mixed reality previsualization

This paper describes the constructing method of consistent action scene data of CG charactors. The proposed method realizes to construct multiple characters' action scenes by merging one actor's action sequences with keeping time and space consistencies between multiple individually recorded or manually-designed action sequences semi-automatically. This technique is planned as a part of "Mixed Reality Pre-visualizing As Pre-Production Tool in Film-making; MR-PreViz" which can streamline the filmmaking process by overlaying CG action data on the real background image in real time.