Twice, again: virentis

Twice, again: Virentis is a virtual world-building installation that uses a fine-tuned large language model and a custom text-to-image network that run locally as part of a generative system. It uses a series of interlinked computational techniques to create new worlds as maps, or perhaps, ideas of maps. The underlying logic of the world-building is based on early personal computer text adventure games and develops the narrative as a relationship between the world itself (Virentis) and a cartographer sent to map the unknown. Image and texts are generated locally, and in real-time, animating across the two screens as an interplay between the two characters. An archive forms from the fragments and new relationships between these pieces are forged—becoming part of a system and incorporated into an overall logic of an emerging world.

Twice, again: Virentis asks a view to consider what it means for a machine to build narrative worlds, in collaboration with an artist/designer? What happens when we bring hand crafted, original, narrative text into play with the “mean”* text generated via large language models, and then set these into play following a set of rules to define a worlds parameters? Where does the line lie between the generated and the generators (both silicon and carbon) and how do we go about reading and interacting with such systems. Can we read them as traditional texts, or do they require a new approach? We also must ask who is(are) the author(s) of such a text and can its creation ever be truly considered a collaboration?