Fairness-Aware Deep Reinforcement Learning for Mitigating Algorithmic Bias in Recommendation Systems

Zhi, Hui

doi:10.1142/S0218001426520099

StoryNode: a node-based visual orchestration framework for automated comic and storyboard generation

2026-08-10

0

PRE

AI

J

Jie Hu

Z

Zhaoli Jiang

Y

Yingfang Zhang *

DOI：10.1007/s00530-026-02567-5

Original

Original request for help

Save

Abstract

En 中文

Maintaining visual and narrative consistency across sequential frames remains a critical bottleneck in digital storytelling. This paper presents StoryNode, an AI-assisted multimedia orchestration framework designed to streamline comic and storyboard generation for novice visual storytellers. By formalizing visual generation as a Directed Acyclic Graph (DAG) orchestration problem, StoryNode integrates Large Language Models (LLMs) for narrative parsing with a context-aware diffusion pipeline and a human-in-the-loop refinement layer. The framework makes three primary contributions: (1) Context-Driven Consistency: Context-Driven Consistency: a fine-tuning-free feature-anchoring workflow that preserves character and style continuity across sequential panels without additional per-character optimization or user-side model-weight updates; (2) End-to-End Workflow Integration: a pipeline connecting LLM scriptwriting, panel generation, style adaptation, and speech-bubble composition; and (3) Accessible Workflow Orchestration: a node-based interface that abstracts complex diffusion parameters into task-oriented controls. We evaluated StoryNode in a counterbalanced within-subjects user study (N = 42) examining what novice users could accomplish with StoryNode and a technically strong IP-Adapter-Plus–ControlNet baseline after the same standardized 10-minute onboarding. The study therefore evaluates short-onboarding accessibility and workflow friction rather than the expert-level capability ceiling of either system. Under this controlled onboarding condition, average production time decreased from 8.7 to 3.4 h, task success increased from 38.8% to 90.5%, and standardized usability measures favored StoryNode (SUS: 82.4; NASA-TLX: 34). These findings suggest that DAG-based orchestration can reduce operational friction and support human-in-the-loop visual storytelling with existing generative models.

Keywords:

Generative AI

Multimodal orchestration

Directed acyclic graph

Character consistency

Human-in-the-loop

Digital storytelling

StoryNode: a node-based visual orchestration framework for automated comic and storyboard generation

Abstract

Journal

Organization

Cited Papers

Citing Papers