Beyond Direct Generation: A Decomposed Approach to Well-Crafted Screenwriting with LLMs

TL;DR

This work tackles the challenge of producing production-ready screenplays with LLMs by decoupling narrative generation from formatting through a two-stage framework called Dual Stage Refinement (DSR). Stage 1 (Outline-to-Novel) generates rich, narrative prose and a CoT reasoning trace, trained via a novel hybrid data-synthesis pipeline that reverse-engineers inputs and forwards to high-quality novel targets; Stage 2 (Novel-to-Screenplay) converts the narrative into professional screenplay format using in-context prompting. Blind evaluations by professional screenwriters show that DSR outperforms strong baselines, with a 75% win rate and about 82.7% of human-level quality, demonstrating that task decoupling and targeted data synthesis yield substantially more coherent, well-structured, and character-driven screenplays. The approach also reveals that model scale, continual pre-training, and CoT reasoning contribute to gains, and that the hybrid synthesis strategy reduces output variance, providing a viable path toward autonomous, production-ready screenplay generation and potentially extrapolating to other structured creative tasks.

Abstract

The screenplay serves as the foundation for television production, defining narrative structure, character development, and dialogue. While Large Language Models (LLMs) show great potential in creative writing, direct end-to-end generation approaches often fail to produce well-crafted screenplays. We argue this failure stems from forcing a single model to simultaneously master two disparate capabilities: creative narrative construction and rigid format adherence. The resulting outputs may mimic superficial style but lack the deep structural integrity and storytelling substance required for professional use. To enable LLMs to generate high-quality screenplays, we introduce Dual-Stage Refinement (DSR), a decomposed framework that decouples creative narrative generation from format conversion. The first stage transforms a brief outline into rich, novel-style prose. The second stage refines this narrative into a professionally formatted screenplay. This separation enables the model to specialize in one distinct capability at each stage. A key challenge in implementing DSR is the scarcity of paired outline-to-novel training data. We address this through hybrid data synthesis: reverse synthesis deconstructs existing screenplays into structured inputs, while forward synthesis leverages these inputs to generate high-quality narrative texts as training targets. Blind evaluations by professional screenwriters show that DSR achieves a 75% win rate against strong baselines like Gemini-2.5-Pro and reaches 82.7% of human-level performance. Our work demonstrates that decomposed generation architecture with tailored data synthesis effectively specializes LLMs in complex creative domains.

Figures (24)

• Figure 1: The proposed two-stage generation pipeline. Stage 1 focuses on creative narrative generation (Outline-to-Novel), while Stage 2 handles structural formatting (Novel-to-Screenplay).
• Figure 2: Workflow of the data synthesizing process. The goal of Step1 is to extract the core structural and narrative elements from a complete screenplay. Step2 aims to understand the "why" behind the screenplay. The objective of Step3 is to generate a new, long-form narrative (a novel) based on the structured information and narrative directives obtained in the previous steps.
• Figure 3: Holistic quality scoring rubric for screenplay evaluation. Each tier represents a quality level with corresponding score ranges and performance criteria.
• Figure 4: Error frequency comparison across models.
• Figure 5: The Prompt Template for Reverse Synthesis of Scene Outline.