New AI method SIC3D enables better control over 3D object generation by combining text descriptions with style images using Gaussian Splatting technology.

SIC3D is a two-stage pipeline that first generates 3D objects from text descriptions, then applies artistic style from reference images

Introduces Variational Stylized Score Distillation (VSSD) loss to capture both global and local texture patterns without conflicts between geometry and appearance

Uses 3D Gaussian Splatting (3DGS) as the core 3D representation for improved detail and control over generated objects

Addresses limitations of text-only generation by enabling image-based style conditioning to reduce texture ambiguity