DramaForge: LLM-based Screenplay Analysis and Adaptation System

DramaForge began ideation in Fall 2024 and was selected for the KAIST Undergraduate Research Program (URP) in Winter–Spring 2025, receiving $3,000 in support.

This work is conducted at the AI Experience Lab (KAIST Industrial Design) under the supervision of Prof. Tak Yeon Lee.

Motivation

Amateur and educational theatre often faces hard constraints (limited cast, short runtime, minimal production resources). Traditional adaptation workflows rely on repeated close readings and iterative restructuring, which is slow and brittle when constraints change. Recent advances in large language models suggest a path toward faster script analysis and constraint-aware adaptation, but practical, production-focused tools for dramatic scripts remain underexplored.

System

DramaForge analyzes scripts across characters, plot, and setting, then proposes constraint-aware adaptations. Examples include: recommending character merges for small casts, prioritizing scene condensation for time limits, and suggesting contemporary re-settings to reduce production cost—while preserving narrative coherence. The system runs as a web application that uses GPT‑4o for parsing text into structured data, performing analysis, and generating adaptation options; the UI presents visualizations alongside editable suggestions.

Future Plan

We plan a user study with university theatre directors to evaluate utility, speed, and creative impact; expand script coverage and constraints; and harden the toolchain into a reliable, production-ready workflow for resource-limited productions.

Repo: GitHub