Narrative optimisation and audience immersive conversion mechanism of cultural tourism short drama driven by generative artificial intelligence

To address the challenges of severe content homogenisation and inefficient audience conversion pathways in cultural tourism short dramas, this study constructs a narrative optimisation framework driven by generative artificial intelligence (GAI).The framework leverages the large-scale, high-quality multilingual dataset for video-and-language research to extract rich audio-visual features.Multimodal large language models are employed to uncover the underlying cultural semantics and dramatic tension embedded in tourism landscapes.From a technical perspective, a cross-modal retrieval mechanism is introduced to achieve precise semantic alignment between landscape imagery and narrative text.Reinforcement learning from human feedback is further applied to dynamically refine the emotional dimensions of the generated narratives.Structural equation modelling is utilised to examine the underlying psychological mechanisms through which audience spatial presence evolves into travel intention.