Generative AI in Film Scene Creation: A Comparison of Platforms

This bachelor's thesis focuses on the use of generative artificial intelligence in the creation of film scenes, specifically on the comparison of two modern platforms: Sora by OpenAI and RunwayML. With the development of these technologies come new possibilities in audiovisual production, but also ethical and professional issues. The motivation for this research stems from the growing presence of AI in film production, which not only raises technological interest but also sparks ethical and professional discussions, especially following the 2023 Hollywood screenwriters' strike. The aim of the thesis was to analyze the current capabilities of these systems in text-to-video generation and to evaluate their applicability in the context of film production. The chosen methodology combined a theoretical review and a comparative analysis of both platforms, including practical testing of four scene types (portrait, landscape, architecture, and fantasy environment). The generated outputs were subsequently assessed through a questionnaire survey involving 52 respondents, including both students and professionals from the film industry. The results showed that Sora was perceived as more visually compelling and better suited for stylized and emotionally driven scenes, while RunwayML demonstrated greater technical stability and practicality in professional settings. Respondents agreed that generative AI currently holds the greatest potential in the pre-production phase, particularly for tasks such as storyboarding, concept visualization, and scene design. However, they also noted that the quality of the outputs is not yet sufficient for direct use without post-production editing. The thesis contributes a practical comparative analysis of two relevant video generation tools, supported by professional feedback. It also highlights limitations such as technical constraints of the platforms, a low number of iterations, and the subjectivity of output evaluation. The findings may serve as guidance for further development of these tools and for decision-making by producers and creators seeking to integrate AI into filmmaking workflows. Future research could explore more advanced model versions and deepen the understanding of their capabilities and limitations in professional production.