Generative AI-Based Night Scene Design of Historic Districts: Value Transition from Basic Lighting to Cultural Narrative

Abstract. As a cultural medium that facilitates the renewal and adaptive reuse of historic neighborhoods, nightscape design is undergoing an evolution from basic illumination to the conveyance of rich cultural narratives. The advent of generative artificial intelligence (AI) in the domain of cultural heritage design has precipitated a paradigm shift, wherein instantaneous engineering has emerged as a pivotal conduit between human intent and AI outcomes. This development has culminated in the enhancement of intelligent visual representations of historic buildings. Generative AI techniques present novel opportunities for conventional lighting design in the context of historic interception. The primary challenge lies in generating nocturnal images that exhibit both aesthetic and cultural depth. This paper explores the potential of generative AI to enhance the efficiency of nightscape design, with a focus on achieving a balance between heritage preservation, contemporary visual language, and cultural significance. The present study proposes a "cultural cue classification system" for transforming daytime imagery into nighttime scenes, based on immediate strategies for cultural heritage. A series of iterative artificial intelligence experiments were conducted using tools such as ChatGPT and Stable Diffusion to evaluate the cultural authenticity, ambience, and detail fidelity of the results, based on the cases of —Nantou Ancient Town and Gankeng Hakka Town in Shenzhen. The findings indicate that culturally embedded cues have a substantial impact on the realism, architectural accuracy, and narrative power of AI-generated images. This accelerated strategy establishes a novel methodological framework for culturally-rich, AI-assisted nighttime narratives.