GeneraBioXR: a novel AI-based framework for the authoring of 3D biophilic models in XR environments

As Virtual Environments (VEs) become more integrated into daily lives, the use of customized virtual spaces grows. Thus, the inclusion of personalized objects in VEs has emerged due to their potential to support user well-being. Biophilic design aims at improving well-being using elements like plants, based on the human connection to nature. However, creating personalized biophilic objects for Extended Reality (XR) experiences remains technically complex. We present GeneraBioXR, a novel framework for the automated generation of 3D biophilic models. It includes all the necessary steps to customize a plant for a text-to-3D generation directly within XR environments. With GeneraBioXR, users without knowledge of modeling and biophilic design can easily customize and generate 3D models of plants. The system gets as input a user textual description and exploits Artificial Intelligence (AI) to create an image, then a multi-view image is generated, followed by mesh reconstruction techniques to produce the final 3D model. Extensive evaluations from a user study revealed a model generation time of about two minutes with 88% generation success rate across 186 plant models, with a high user satisfaction with the generated 3D models. Users found the system easy to use, with a low workload, and perceived it to be more beneficial compared to other existing authoring methods.