StoryState: Agent-Based State Control for Consistent and Editable Storybooks

2026-02-01

AI Summary

This work addresses the limitations of existing text-to-image children’s book generation methods, which lack explicit story state representations, resulting in coarse editing granularity and poor cross-page visual consistency. The authors propose a multi-agent coordination framework that, without requiring training of text-to-image models, introduces a structured story state—comprising a character table, global setting, and per-page scene constraints—maintained collaboratively by multiple LLM agents to generate prompts aligned with the 1Prompt1Story paradigm. By integrating explicit story states with LLM agents for the first time, this approach enables fine-grained, editable, and cross-page consistent storybook generation, exhibiting model-agnosticism and efficient interactivity. Experiments demonstrate significant improvements in consistency during multi-page editing tasks, along with reduced unintended modifications, fewer interaction rounds, and shorter editing time, achieving performance comparable to Gemini Storybook’s one-click generation.

Technology Category

• Multiagent Systems: Coordination and Collaboration
• Humans and AI: Game Design — Procedural Content Generation & Storytelling
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Large multimodal models have enabled one-click storybook generation, where users provide a short description and receive a multi-page illustrated story. However, the underlying story state, such as characters, world settings, and page-level objects, remains implicit, making edits coarse-grained and often breaking visual consistency. We present StoryState, an agent-based orchestration layer that introduces an explicit and editable story state on top of training-free text-to-image generation. StoryState represents each story as a structured object composed of a character sheet, global settings, and per-page scene constraints, and employs a small set of LLM agents to maintain this state and derive 1Prompt1Story-style prompts for generation and editing. Operating purely through prompts, StoryState is model-agnostic and compatible with diverse generation backends. System-level experiments on multi-page editing tasks show that StoryState enables localized page edits, improves cross-page consistency, and reduces unintended changes, interaction turns, and editing time compared to 1Prompt1Story, while approaching the one-shot consistency of Gemini Storybook. Code is available at https://github.com/YuZhenyuLindy/StoryState

Problem

• story state
• visual consistency
• editable storybooks
• multimodal generation
• coarse-grained editing

Innovation

• agent-based orchestration
• explicit story state
• editable storybooks
• prompt engineering
• cross-page consistency