Writing Films with AI – Storytelling with LLMs

The project “Writing Films with AI – Storytelling with LLMs” is a scientific-artistic investigation into the evolving landscape of creative collaboration, specifically focusing on the process of writing a feature film screenplay with a Large Language Model (LLM). This research moves beyond the speculative question of whether machines can be creative and instead provides a practical, documented exploration of how human-machine interaction currently functions in a narrative context, analyzing the implications of this technology for the screenwriting profession and the creative industries at large.

The central thesis that emerges from this research is that the success of a human-AI screenwriting collaboration is not determined by the AI’s autonomy, but is directly proportional to the quality of the human collaborator’s creative guidance, critical judgment, and emotional input. The study posits that LLMs function as powerful augmentative tools—not as authors—whose output of clichés or innovation is contingent on the human’s ability to act as a visionary curator, editor, and the essential source of lived experience.

To build this argument, the project was guided by an integrated set of research questions. It began by investigating how human and machine creativity can engage in a fruitful exchange to produce innovative film narratives. This was not explored theoretically, but through a hands-on artistic research methodology: a writer’s room scenario where a human writer and the LLM collaborated to develop, write, and refine a complete, feature-length screenplay.

This practical experiment was designed to reveal the specific benefits and limitations of using LLMs in the screenwriting process. The findings demonstrate that the LLM is highly effective for brainstorming, generating plot variations, overcoming writer’s block, and structuring scenes based on established narrative formulas. However, the research also highlights significant limitations, including the model’s tendency to produce generic, clichéd content and its fundamental inability to generate true emotional depth, which the thesis argues can only stem from the nuances of human consciousness and lived experience. 

Ultimately, the project addresses the wider implications of AI-generated content on the creative industry and how to ensure LLMs serve as beneficial collaborators rather than replacing human creativity. The research concludes that the writer’s role does not disappear but evolves. The human becomes the keeper of the vision, taste, and thematic coherence, guiding the AI tool away from derivative content and towards a more compelling narrative. The project proposes a collaborative model that leverages the strengths of both partners: the machine’s speed and processing power with the human’s emotional intelligence and artistic judgment. It suggests that the future of storytelling with these tools lies in a symbiotic partnership that enhances, rather than replaces, the indispensable human element of creativity.