We present a VR-based pipeline for immersive exploration of cultural heritage images using camera-guided video diffusion and 3D Gaussian Splatting (3DGS). Starting from restored historical photographs, we synthesize novel views via diffusion models and reconstruct spatial scenes with 3DGS, enabling users to experience heritage sites as interactive environments. Integrated into a virtual gallery, our system allows seamless transitions between 2D images and 3D spatial representations. A formative evaluation with 14 participants highlights the potential of generative pipelines for cultural heritage VR, while also identifying challenges related to visual artifacts. Finally, we reflect on the promises and challenges of the proposed pipeline, especially regarding content stability and automation.
Kobets et al. (Wed,) studied this question.