Computational methods to analyze and build creative contents are important issues to enable collaboration between creators and softwares with artificial intelligence. Despite current advances of machine learning, there is a few data of common creative contents with the process of creating stories. In order to solve the problem,

I focus on four-scene comics, which is one of the multi-modal creative contents, as the first example.

The four scene comics have a clear and simple structure; the length of a story is limited, the shape of each scene is always a rectangle, and the size of a scene is consistent throughout the entire comic book.

In this paper, I construct four scene-comic story dataset with plots and layers from both view points of several professional creators and engineering researcher.