Plug-and-Play Dramaturge: A Divide-and-Conquer Approach for Iterative Narrative Script Refinement via Collaborative LLM Agents

TL;DR

The paper addresses the difficulty of achieving high-quality long-form narratives with single-pass LLM generation. It introduces Dramaturge, a plug-and-play, divide-and-conquer framework that iteratively refines scripts via Global Review, Scene-level Review, and Hierarchical Coordinated Revision, guided by task- and feature-specific agents. Across a 50-script dataset spanning five genres, Dramaturge significantly outperforms strong baselines at both script-level quality and scene-level detail, with substantial gains and robust ablation analyses demonstrating the effectiveness of the hierarchical, multi-agent approach. The results demonstrate that coarse-to-fine, top-down information flow with integrated quality control yields deeper narrative coherence, richer character development, and more immersive scene presentation, offering a practical path toward AI-assisted, high-quality scriptwriting.

Abstract

Although LLMs have been widely adopted for creative content generation, a single-pass process often struggles to produce high-quality long narratives. How to effectively revise and improve long narrative scripts like scriptwriters remains a significant challenge, as it demands a comprehensive understanding of the entire context to identify global structural issues and local detailed flaws, as well as coordinating revisions at multiple granularities and locations. Direct modifications by LLMs typically introduce inconsistencies between local edits and the overall narrative requirements. To address these issues, we propose Dramaturge, a task and feature oriented divide-and-conquer approach powered by hierarchical multiple LLM agents. It consists of a Global Review stage to grasp the overall storyline and structural issues, a Scene-level Review stage to pinpoint detailed scene and sentence flaws, and a Hierarchical Coordinated Revision stage that coordinates and integrates structural and detailed improvements throughout the script. The top-down task flow ensures that high-level strategies guide local modifications, maintaining contextual consistency. The review and revision workflow follows a coarse-to-fine iterative process, continuing through multiple rounds until no further substantive improvements can be made. Comprehensive experiments show that Dramaturge significantly outperforms all baselines in terms of script-level overall quality and scene-level details. Our approach is plug-and-play and can be easily integrated into existing methods to improve the generated scripts.

Figures (9)

• Figure 1: Our Dramaturge is inspired by the human scriptwriting process and performs Global Review, Scene-level Review, and Hierarchical Coordinated Revision to iteratively refine narrative scripts via a task and feature oriented divide-and-conquer strategy.
• Figure 2: The Architecture of Dramaturge. A task and feature oriented divide-and-conquer strategy is adopted, leveraging collaborative LLM agents to perform coarse-to-fine iterative refinement, while coordinating local edits and global structural adjustments and maintaining contextual consistency.
• Figure 3: Enhancement in character development. Dramaturge introduces internal conflict and a subplot for Ron, transforming him from a sidekick into a more well-rounded character through foreshadowing and payoff.
• Figure 4: Enhancement in narrative structure. Dramaturge refines storyline to link external conflicts with internal character arcs, introduces subplots, and externalizes abstract themes into concrete plots, creating a more resonant story.
• Figure 5: Enhancement of scene presentation. Dramaturge introduces atmospheric intensification and character-environment integration to create a more immersive and psychologically impactful scene, which mirrors characters’ internal states and provides narrative foreshadowing.