DiaryPlay: AI-Assisted Creation of Interactive Story Vignettes for Everyday Storytelling

Paper Title

DiaryPlay: AI-Assisted Creation of Interactive Story Vignettes for Everyday Storytelling

Publication Info

• Topic area: AI-assisted tools for interactive storytelling in casual contexts.
• Keywords: AI-assisted authoring, interactive vignettes, everyday storytelling, large language models, NPC behaviors, branching narratives, human-AI co-creation, social media storytelling, lightweight authoring, immersive experiences.

Background and Problem

• Problem / challenge: Existing tools for creating interactive vignettes are resource-intensive, technically demanding, and time-consuming, making them unsuitable for the immediacy required in everyday storytelling.
• Significance: Interactive vignettes can enrich everyday storytelling by providing immersive, participatory experiences, but their complexity limits adoption by casual storytellers.
• Motivation and related work: Prior research has explored interactive storytelling tools and generative AI systems, but these focus on professional or game design contexts rather than lightweight everyday use. No existing solution balances authoring simplicity with adaptive branching narratives for casual storytelling.

Solution

• Proposed approach: DiaryPlay, an AI-assisted authoring system that transforms text stories into interactive vignettes, enabling casual authors to create immersive storytelling experiences with minimal effort.
• Novelty:
  1. Introduction of the Controlled Divergence (CD) Module, which enables adaptive NPC behaviors and branching narratives while preserving the author’s intended storyline.
  2. Automated extraction and refinement of interactive vignette elements (environment, characters, events) from text inputs.
  3. Lightweight authoring workflow tailored for everyday storytelling practices.
  4. Empirical validation of the system’s ability to generate believable NPC activities and support engaging viewing experiences.
• Procedure and key techniques:
  • Authors input a text story, which the system processes to extract environment layouts, character profiles, and event sequences.
  • Authors refine these elements using an intuitive interface, with AI assistance for tasks like persona building and environment editing.
  • During viewing, the CD Module dynamically plans NPC activities and reacts to player interactions using a branch-and-bottleneck narrative structure.

Results

• Concrete findings:
  • NPC activities generated by the CD Module ranked significantly better in believability than baseline methods (μ = 2.37 vs. μ = 4.00; p < 0.01) and were on par with human-authored activities (μ = 2.58; p = 0.71).
  • Participants completed interactive vignette authoring in 12–19 minutes (μ = 16), meeting expectations for lightweight effort.
  • User study participants described the viewing experience as engaging and immersive, with NPCs behaving believably and supporting divergent player interactions.
• Advantage over baselines:
  • The CD Module outperformed baseline methods in generating believable NPC activities and demonstrated the importance of combining persona and storyline properties.
  • DiaryPlay reduced authoring complexity compared to traditional tools like Unity and RPGMaker.
• Experiments / evaluation:
  • Preliminary technical evaluation with 16 participants assessed NPC activity believability across five conditions.
  • User study with 16 participants evaluated authoring and viewing experiences, focusing on system usability and narrative fidelity.
• Limitations and future work:
  • Limited to single-location, event-centric stories; future work could expand to multi-location narratives and richer emotional modeling.
  • Controlled lab setting and culturally homogeneous participant pool; longer-term, cross-cultural studies are needed.
  • Current system supports only three characters and lacks mobile device optimization.

Summary

DiaryPlay introduces an AI-assisted system for creating interactive vignettes tailored to everyday storytelling. By automating the extraction of narrative elements and enabling adaptive NPC behaviors through the Controlled Divergence Module, DiaryPlay reduces authoring effort while preserving narrative fidelity. Empirical evaluations demonstrate its ability to generate believable NPC activities and deliver engaging viewing experiences. While currently limited to simple, single-location stories, DiaryPlay offers a promising foundation for integrating interactive storytelling into social media and other casual contexts. Future work could expand its scope, enhance mobile usability, and explore applications beyond everyday storytelling.