Generative AI for Interactive Storytelling

Overview

Generative AI models have shown remarkable capabilities in creating diverse content, including text, images, and voices. However, Large Language Models often struggle to maintain thematic consistency and structure when used for narrative generation, especially when creating longer and more complex stories. When generative AI aims to produce high-quality narratives, having mechanisms to guide the creative process and ensure thematic consistency is crucial.

This project explores new methods to guide the generation of narratives using AI, by incorporating narrative patterns, genre structures, and semiotic relations, to improve the overall quality and coherence of the generated stories. The research addresses fundamental questions about how to structure AI-generated narratives while maintaining creative flexibility and supporting interactive storytelling applications.

Generative AI for Interactive Storytelling

Research Challenges

How can we maintain thematic consistency in AI-generated narratives across multiple scenes and story arcs?
What narrative structures and patterns from folklore and literary traditions can effectively guide generative AI systems?
How can semiotic operations enable the systematic reconstruction of narratives while preserving coherence?
What role can genre structures play in constraining and guiding the creative process of large language models?

Approach

This research explores multiple complementary approaches to guide and structure AI-generated narratives. The work investigates how different types of constraints and guidance mechanisms can improve the coherence and quality of stories produced by large language models.

One direction examines how semiotic theory can provide systematic operations for narrative reconstruction, enabling the transformation of existing stories through combination, imitation, expansion, and reversal. Another direction investigates how narrative patterns derived from folklore traditions and the structural characteristics of established genres can serve as effective constraints for the generation process.

The research also explores how multimodal capabilities of contemporary AI systems can be used for narrative generation, including generating stories from visual inputs, transforming news articles into fictional narratives, and adapting story content to different player types and preferences. Additionally, the project examines how character-driven approaches and the extraction of narrative traits from existing works can inform generation methods.

Throughout this research, interactive prototypes are developed to validate and demonstrate different methods, allowing exploration of how human-AI collaboration can support creative storytelling while maintaining narrative quality and coherence.

Key Contributions

Semiotic Reconstruction Framework

Developed an approach to interactive narrative generation based on semiotic relations, enabling the creation of new narratives from existing ones through systematic operations along syntagmatic, paradigmatic, and meronymic axes.

Pattern-Guided Story Generation

Introduced methods using narrative patterns from folklore traditions and genre structures to guide LLM-based story composition, ensuring thematic consistency while maintaining creative flexibility.

Multimodal Narrative Generation

Developed methods for generating narratives from visual inputs using multimodal large language models, including player-driven stories based on game images and narrative generation from image sequences.

News-to-Story Transformation

Created techniques for transforming factual news content into fictional narratives using genre-based, structured, and data-driven retelling approaches with large language models.

Character and Genre Analysis

Developed AI-driven methods for characterizing narrative elements such as detective investigative methods and for analyzing gender bias in LLM-generated narratives, advancing understanding of how AI systems construct character-driven stories.

Interactive Prototypes

Explore the interactive prototypes developed as part of this research. Each tool demonstrates different approaches to AI-powered narrative generation and is freely available online:

ChatGeppetto

Semiotic narrative reconstruction tool that creates new stories from existing ones (movies and books) through combination, imitation, expansion, and reversal operations.

Try it →

PatternTeller

Pattern-oriented story composition using narrative patterns and genre structures to guide thematically consistent story generation.

Try it →

ImageTeller

Image-driven narrative generation that creates stories from single images or sequences, exploring multimodal AI capabilities for visual storytelling.

Try it →

NewsReteller

Transforms news articles into fictional narratives using genre-based, structured, and data-driven retelling approaches with large language models.

Try it →

Related Publications

Irene C.E. van Blerck; Edirlei Soares de Lima; Margot M.E. Neggers; Toon Calders. Unveiling Gender Bias in LLM-generated Hero and Heroine Narratives. Entertainment Computing, Volume 55, September 2025, 100972. [DOI]
Edirlei Soares de Lima; Marco Antonio Casanova; Bruno Feijo; Antonio Furtado. From News to Stories via an AI-supported Retelling Process. 24th IFIP International Conference on Entertainment Computing (ICEC 2025), Tokyo, Japan, 2025. [DOI] [PDF]
Irene C.E. van Blerck; Edirlei Soares de Lima. Exploring Gender Bias in LLM-Generated Narratives. Workshop at the 24th IFIP International Conference on Entertainment Computing (ICEC 2025), Tokyo, Japan, 2025. [DOI] [PDF]
Edirlei Soares de Lima; Marco A. Casanova; Bruno Feijó; Antonio L. Furtado. Characterizing the Investigative Methods of Fictional Detectives with Large Language Models. arXiv:2505.07601 [cs.CL], May 2025. [DOI]
Edirlei Soares de Lima; Margot M.E. Neggers; Bruno Feijó; Marco A. Casanova; Antonio L. Furtado. An AI-powered Approach to the Semiotic Reconstruction of Narratives. Entertainment Computing, Volume 52, January 2025, 100810. [DOI]
Edirlei Soares de Lima; Margot M.E. Neggers; Marco A. Casanova; Bruno Feijó; Antonio L. Furtado. A Pattern-oriented AI-powered Approach to Story Composition. 23rd IFIP International Conference on Entertainment Computing (ICEC 2024), Manaus, Brazil, 2024. [DOI] [PDF]
🏆 Best Paper Award
Edirlei Soares de Lima; Margot M.E. Neggers; Marco A. Casanova; Antonio L. Furtado. From Images to Stories: Exploring Player-Driven Narratives in Games. 14th International Conference on Videogame Sciences and Arts (VIDEOJOGOS 2024), Leiria, Portugal, 2024. [DOI] [PDF]
Edirlei Soares de Lima; Marco A. Casanova; Antonio L. Furtado. Imagining from Images with an AI Storytelling Tool. arXiv:2408.11517 [cs.CL], August 2024. [DOI]
Edirlei Soares de Lima, Margot M. E. Neggers, Antonio L. Furtado. Multigenre AI-powered Story Composition. arXiv:2405.06685 [cs.CL], May 2024. [DOI]
Victor Schetinger; Dafne Reis Pedroso da Silva; Sara Di Bartolomeo; Edirlei Soares de Lima; Christofer Meinecke; Rudolf Rosa. Macunaíma, AI Parrot, Solves Crimes in Prague: Towards Pattern Visualization in Generative AI Model Narratives. Revista GEMInIS, vol. 14 (3), pp. 21-37, 2024. [DOI] [PDF]
Edirlei Soares de Lima; Marco A. Casanova; Bruno Feijo; Antonio L. Furtado. Semiotic Structuring in Movie Narrative Generation. 22nd IFIP International Conference on Entertainment Computing (ICEC 2023), Bologna, Italy. Lecture Notes in Computer Science, vol. 14455, pp. 161-175, 2023. [DOI] [PDF]
Edirlei Soares de Lima; Bruno Feijo; Marco A. Casanova; Antonio L. Furtado. ChatGeppetto - an AI-powered Storyteller. Proceedings of the 22nd Brazilian Symposium on Games and Digital Entertainment (SBGames '23), Rio Grande, Brazil, pp. 28-37, 2023. [DOI] [PDF]
🏆 Best Paper Award
Victor Schetinger, Sara Di Bartolomeo, Edirlei Soares de Lima, Christofer Meinecke, Rudolf Rosa. n Walks in the Fictional Woods. arXiv:2308.06266 [cs.HC], August 2023. [DOI]