Stella generates a story based on an ’emotion curve’ selected by a human. The user navigates Stella’s GUI to set constraints on the story, such as the story type (e.g. love, murder), number of characters in the story, the length of the story, and the story’s emotion curve. The emotion curve controls the narrative arc of the story. Stella then writes the story according to the constraints. The user can re-adjust the constraints and generate a new story.
Stella’s algorithm ‘traverses a generative search space in which new stories are generated by the iterative application of domain rules on a initial story, therefore including new parts of the story on each step. Only addition rules are permitted, which eases the rule definition process and allows a better optimization of the generation engine.’