Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

HoLLMwood: Unleashing the Creativity of Large Language Models in Screenwriting via Role Playing

Jing Chen, Xinyu Zhu, Cheng Yang, Chufan Shi, Yadong Xi, Yuxiang Zhang, Junjie Wang, Jiashu Pu, Rongsheng Zhang, Yujiu Yang, Tian Feng

TL;DR

HoLLMwood presents a fully automated, role-based framework for screenwriting that assigns LLMs to Writer, Editor, and Actors to emulate the human creative process. By structuring the workflow into Plot Planning, Story Expansion, and Screenplay Generation, and by introducing a feedback loop and role-playing interactions, the approach achieves substantial gains in coherence, relevance, and especially interestingness over strong baselines. Through GPT-4–based pairwise evaluations across six genres, the work shows that explicit collaboration and human-like roles unlock higher-quality scripts, with ablations confirming the value of both feedback-revision and role-playing. The study highlights practical considerations, such as screenplay length and the challenges of open-weight models, and suggests that this modular, automated pipeline can democratize high-quality screenwriting while leaving room for future improvements and open-source development.

Abstract

Generative AI has demonstrated unprecedented creativity in the field of computer vision, yet such phenomena have not been observed in natural language processing. In particular, large language models (LLMs) can hardly produce written works at the level of human experts due to the extremely high complexity of literature writing. In this paper, we present HoLLMwood, an automated framework for unleashing the creativity of LLMs and exploring their potential in screenwriting, which is a highly demanding task. Mimicking the human creative process, we assign LLMs to different roles involved in the real-world scenario. In addition to the common practice of treating LLMs as ${Writer}$, we also apply LLMs as ${Editor}$, who is responsible for providing feedback and revision advice to ${Writer}$. Besides, to enrich the characters and deepen the plots, we introduce a role-playing mechanism and adopt LLMs as ${Actors}$ that can communicate and interact with each other. Evaluations on automatically generated screenplays show that HoLLMwood substantially outperforms strong baselines in terms of coherence, relevance, interestingness and overall quality.

HoLLMwood: Unleashing the Creativity of Large Language Models in Screenwriting via Role Playing

TL;DR

HoLLMwood presents a fully automated, role-based framework for screenwriting that assigns LLMs to Writer, Editor, and Actors to emulate the human creative process. By structuring the workflow into Plot Planning, Story Expansion, and Screenplay Generation, and by introducing a feedback loop and role-playing interactions, the approach achieves substantial gains in coherence, relevance, and especially interestingness over strong baselines. Through GPT-4–based pairwise evaluations across six genres, the work shows that explicit collaboration and human-like roles unlock higher-quality scripts, with ablations confirming the value of both feedback-revision and role-playing. The study highlights practical considerations, such as screenplay length and the challenges of open-weight models, and suggests that this modular, automated pipeline can democratize high-quality screenwriting while leaving room for future improvements and open-source development.

Abstract

Generative AI has demonstrated unprecedented creativity in the field of computer vision, yet such phenomena have not been observed in natural language processing. In particular, large language models (LLMs) can hardly produce written works at the level of human experts due to the extremely high complexity of literature writing. In this paper, we present HoLLMwood, an automated framework for unleashing the creativity of LLMs and exploring their potential in screenwriting, which is a highly demanding task. Mimicking the human creative process, we assign LLMs to different roles involved in the real-world scenario. In addition to the common practice of treating LLMs as , we also apply LLMs as , who is responsible for providing feedback and revision advice to . Besides, to enrich the characters and deepen the plots, we introduce a role-playing mechanism and adopt LLMs as that can communicate and interact with each other. Evaluations on automatically generated screenplays show that HoLLMwood substantially outperforms strong baselines in terms of coherence, relevance, interestingness and overall quality.
Paper Structure (25 sections, 4 figures, 10 tables)

This paper contains 25 sections, 4 figures, 10 tables.

Table of Contents

  1. Introduction
  2. HoLLMwood
  3. Plot Planning
  4. Story Expansion
  5. Screenplay Generation
  6. Experiments
  7. Experimental Setup
  8. Dataset
  9. Baselines
  10. Evaluation
  11. Results
  12. Ablation Study
  13. Impact of Feedback-Revision
  14. Impact of Role-Playing
  15. Analysis
  16. ...and 10 more sections

Figures (4)

  • Figure 1: An overview of HoLLMwood for automatic screenwriting.
  • Figure 2: An example of plot planning. Highlighted texts refer to the feedback parts proposed by the editor.
  • Figure 3: An example of story expansion. Highlighted texts refer to the parts expanded from the given input.
  • Figure 4: An example of screenplay generation with the script draft in the left, the screenplay with role playing in the middle and the actor context in the right.