VcEdit is a novel method for view-consistent 3D editing using Gaussian splatting. It addresses the challenge of maintaining consistency across multiple views during 3D scene editing. The method leverages two key modules: the Cross-attention Consistency Module (CCM) and the Editing Consistency Module (ECM). The CCM enhances view consistency by consolidating multi-view cross-attention maps, ensuring coherent editing guidance across different perspectives. On the other hand, the ECM enforces consistency in the edited 3D geometry and texture, facilitating accurate and coherent modifications.

VcEdit involves a multi-stage process, including initial 3D geometry generation, multi-view consistent guidance consolidation, and iterative refinement of the 3D geometry and texture. The method utilizes a training set to steer 3D geometry and texture modifications iteratively, ensuring that the edited 3D scenes maintain view consistency and accurately reflect the user's prompts. VcEdit demonstrates its adaptability in handling a wide range of challenging scenes and prompts, showcasing its competency in delivering high-quality edits across diverse editing tasks, including face, object, and large-scale scene transformations.