TaleFrame: An Interactive Story Generation System with Fine-Grained Control and Large Language Models

2025-12-02

AI Summary

Current story generation systems suffer from imprecise user intent realization due to ambiguous inputs and coarse-grained control. To address this, we propose a structured human-AI collaborative story generation framework: stories are decomposed into four orthogonal units—entities, events, relations, and outlines—and generated via a novel JSON2Story paradigm. We fine-tune Llama locally on a curated, preference-aligned dataset of 9,851 structured samples derived from TinyStories. The system integrates visual interaction mechanisms—including drag-and-drop and node-link editing—to enable fine-grained intent modeling and iterative refinement. Experimental evaluation and user studies demonstrate statistically significant improvements across seven quantitative and qualitative dimensions (creativity, logical consistency, coherence, etc.), markedly enhancing narrative controllability and output quality. Our work establishes a new paradigm for interpretable, evaluable, and interactive story generation.

Technology Category

• Humans and AI: Game Design — Procedural Content Generation & Storytelling
• Natural Language Processing: Generation
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

With the advancement of natural language generation (NLG) technologies, creative story generation systems have gained increasing attention. However, current systems often fail to accurately translate user intent into satisfactory story outputs due to a lack of fine-grained control and unclear input specifications, limiting their applicability. To address this, we propose TaleFrame, a system that combines large language models (LLMs) with human-computer interaction (HCI) to generate stories through structured information, enabling precise control over the generation process. The innovation of TaleFrame lies in decomposing the story structure into four basic units: entities, events, relationships, and story outline. We leverage the Tinystories dataset, parsing and constructing a preference dataset consisting of 9,851 JSON-formatted entries, which is then used to fine-tune a local Llama model. By employing this JSON2Story approach, structured data is transformed into coherent stories. TaleFrame also offers an intuitive interface that supports users in creating and editing entities and events and generates stories through the structured framework. Users can control these units through simple interactions (e.g., drag-and-drop, attach, and connect), thus influencing the details and progression of the story. The generated stories can be evaluated across seven dimensions (e.g., creativity, structural integrity), with the system providing suggestions for refinement based on these evaluations. Users can iteratively adjust the story until a satisfactory result is achieved. Finally, we conduct quantitative evaluation and user studies that demonstrate the usefulness of TaleFrame. Dataset available at https://huggingface.co/datasets/guodaosun/tale-frame.

Problem

• Addresses lack of fine-grained control in story generation systems
• Translates user intent into structured story units for precise output
• Enables interactive editing and evaluation for iterative story refinement

Innovation

• Combines LLMs with structured story units for control
• Fine-tunes local Llama model using parsed JSON dataset
• Provides interactive interface for editing and iterative refinement