Section: New Results

Interactive Stroytelling

Participants : Marc Christie [contact] , Hui-Yin Wu.

In 2013, within the Inria Associate Team FORMOSA (see ), we have proposed a framework for the creation of parametrable and personalized stories in interactive storytelling. In any kind of storytelling, the success of the story relies both on the intricate plot design and control of the author as well as the emotional feedback of the user. With the assistance of computing algorithms combined with the maturing understanding of narrative structures, it is possible for interactive stories to create a more personalized, engaging, and well-controlled narrative content to users than traditional linear narrative. And with the emergence of new storytelling technologies, critical issues concern the creation of such complex narratives in virtual 3D environments, and the coherent simulation of these interactive narratives.

In the framework we proposed, the author can specify characteristics on the story structure and fragments (pieces of story) in order to generate variations of interactive stories. The characteristics we consider are genre, story complexity, and Chatman's modes of plot (eg a good hero fails). The story generation model we devised combines a branching story structure with a three-step graph traversal algorithm that filters and recombines story fragments from the characteristics, generating a high-level interactive script that satisfies all authorial constraints, and provides sufficient abstraction from the technical implementation. The script is then simulated in a real-time storytelling system, featuring autonomous characters and automatic camera control. The work has been presented as a short paper in the CASA conference [30] .

We then extended this approach to handle temporal aspects of discourse in stories (i.e. how to temporally rearrange fragments of a story while maintaining consistency and logic whatever the user's choices). By rewriting our graph traversal algorithm (which filters inconsistent branches, and propagates constraints along the branches), and performing the graph traversal on each choice selected by the user, we enable the simulation of consistent temporal variations in stories. This typically allows the creation of flashbacks, flashforwards, parallel and embedded stories. Early results have been presented as a poster as Motion in Games 2013 [21] .