🤖 AI Summary
This work addresses the inconsistent modeling quality of current text-to-3D and image-to-3D generative tools. We present the first systematic, critical evaluation of mainstream online generators—including diffusion-based, NeRF-based, and implicit surface reconstruction methods—assessing their real-world output performance. Using a multi-prompt diversity benchmark coupled with a hybrid human-and-automated evaluation framework, we identify pervasive geometric distortions in complex topologies and fine-grained structures, and uncover key bottlenecks at the intersection of prompt engineering, geometric fidelity, and semantic consistency. Our contributions include actionable prompt design principles and a standardized set of quality evaluation metrics. These provide an empirical benchmark for applications such as digital twins and advance the development of next-generation generative 3D modeling technologies.
📝 Abstract
In the world of technology and AI, digital models play an important role in our lives and are an essential part of the digital twins of real-world objects. They can be created by designers, artists, or game developers using spline curves and surfaces, meshes, and voxels, but making such models is too time-consuming. With the growth of AI tools, there is interest in the automated generation of 3D models, such as generative design approaches, which can save creators valuable time. This paper reviews several online 3D model generators and critically analyses the results, hoping to see higher-quality results from different prompts.