The Generative Audiovisual Narrative System

This paper documents the results of a research project that deals with the application of an artificial life (AL) approach to creating a system of cinematographic narration. This project investigates the possibility of producing an autonomous cinematographic narration system, in which meaning results from a kind of hypermontage, conditioned by genetic algorithms. The theoretical thesis lead to the design and construction of a generative system for the synthesis of audiovisual narratives, in which a genetic algorithm manages the collection of video fragments, that have been parameterized and stored in a database. The genetic algorithm’s fitness criteria are being defined by the viewer, via the system’s interface. During the presentation of this paper, the system and its audiovisual results will be demonstrated. This paper investigates whether individual video fragments belonging to a database may be linked into numerous different successions in order to satisfy fitness criteria defined by the user. The aim is to create optimum sequences in accordance to specific requirements, instead of coming up with a closed predetermined unique sequence, as it’s traditionally done by directors. Fragmented videos as micro-narrations are composed to a greater whole which is unique, since the system has never given the same result twice. Taking the phenomenon of semantic montage as an axiom, the viewer attributes causality relationships to the succession of these fragmented micro-narrations which are seamlessly integrated in the sequence. A traditional work of art may be read in multiple ways, which depend on the subjectivity and arbitrariness of the spectator-reader. In the case of AL, the multiplicity of the artwork mainly depends on the manner in which the artifact is being written. Having the same structural elements as a starting point may result in different readings. Having the same data as a starting point, may lead to different versions of the same work of art