This paper presents a method for generating player-driven narratives from visual inputs by exploring the visual analysis capabilities of multimodal large language models. By employing Bartle’s taxonomy of player types—Achievers, Explorers, Socializers, and Killers—our method creates stories that are tailored to different player characteristics. We conducted a fourfold experiment using a set of images extracted from a well-known game, generating distinct narratives for each player type that are aligned with the visual elements of the input images and specific player motivations. By adjusting narrative elements to emphasize achievement for Achievers, exploration for Explorers, social connections for Socializers, and competition for Killers, our system produced stories that adhere to established narratology principles while resonating with the characteristics of each player type. This approach can serve as a helping tool for game designers, offering new insights into how players might engage with game worlds through personalized image-driven narratives.
