Editable 3D

Editable 3D research focuses on creating and manipulating three-dimensional models using intuitive methods, aiming for user-friendly interfaces and high-fidelity results. Current approaches leverage implicit representations like Gaussian splatting and neural radiance fields (NeRFs), often combined with diffusion models and visual transformers, to enable editing based on 2D inputs (e.g., sketches or images) or text prompts. This allows for precise control over object shape, texture, and even animation, with applications ranging from 3D modeling software to virtual and augmented reality experiences. The development of part-aware models further enhances the ability to edit specific regions of a 3D object independently.