A Study of Cultural Differences among Supernatural Beings in Horror Movies and Generative AI Image Generation - Comparison of Visual Expressions in Korean, Japanese, and American Horror Films, Focusing on DALL · E

This study aims to compare and analyze the visual presentation of supernatural beings in Korean, Japanese, and American horror films and to examine the possibilities and limitations of cultural visual expression through generative AI image experiments based on the comparison. Horror films are a genre that audio-visually represents the anxiety and oppression of each society, and supernatural beings are constructed differently in each cultural sphere according to the emotional structure and image strategy. However, generative-AI-based image generation technology reveals the problem of omitting the unique sensibility and context of a culture or reducing it to a universal iconography. Therefore, this study conducted an analysis using six cases (Korean “Pamyo” and “Gonjiam,” Japanese “Under the Black Water” and “The Ring,” and American “Lights Out” and “The Conjuring”) This study selected six commercially successful films released after 2022, all of which depict supernatural beings as a central narrative element. Through a literature review and scene analysis, we structured the shape, lighting, space, gaze, and sensory composition elements in the movies from each country, designed prompts based on these, and then conducted an image generation experiment using the ChatGPT-4o-based DALL·E model. It was found that the generated images reflected some visual icons but showed limitations in implementing temporality, psychological rhythm, and sensory cumulative structures. In particular, with regard to East Asia, the problems of the repeated generation of clichéd icons and the omission of any emotional context were confirmed. However, cultural visual elements at the shape level were confirmed to have a certain level of reproducibility, with the possibility of generating advanced images reflecting cultural sensibilities raised through future sequence-based design and emotion-based prompt supplementation. This study has research value as an attempt to structure the visual strategy of horror movies and analytically approach the possibilities and limitations of generative AI image reproduction, and we expect that this paper will have a positive impact on future research on cultural foundations in movies.