SplatFlow: Multi-View Rectified Flow Model for 3D Gaussian Splatting Synthesis
This provides a more streamlined approach for content creators in 3D graphics by unifying generation and editing, though it appears incremental as it builds on existing 3DGS and rectified flow methods.
The paper tackles the lack of a unified framework for text-based generation and editing of 3D scenes using 3D Gaussian Splatting, introducing SplatFlow which integrates a multi-view rectified flow model and a Gaussian Splatting Decoder to enable direct 3DGS synthesis and editing across tasks like object editing and novel view synthesis.
Text-based generation and editing of 3D scenes hold significant potential for streamlining content creation through intuitive user interactions. While recent advances leverage 3D Gaussian Splatting (3DGS) for high-fidelity and real-time rendering, existing methods are often specialized and task-focused, lacking a unified framework for both generation and editing. In this paper, we introduce SplatFlow, a comprehensive framework that addresses this gap by enabling direct 3DGS generation and editing. SplatFlow comprises two main components: a multi-view rectified flow (RF) model and a Gaussian Splatting Decoder (GSDecoder). The multi-view RF model operates in latent space, generating multi-view images, depths, and camera poses simultaneously, conditioned on text prompts, thus addressing challenges like diverse scene scales and complex camera trajectories in real-world settings. Then, the GSDecoder efficiently translates these latent outputs into 3DGS representations through a feed-forward 3DGS method. Leveraging training-free inversion and inpainting techniques, SplatFlow enables seamless 3DGS editing and supports a broad range of 3D tasks-including object editing, novel view synthesis, and camera pose estimation-within a unified framework without requiring additional complex pipelines. We validate SplatFlow's capabilities on the MVImgNet and DL3DV-7K datasets, demonstrating its versatility and effectiveness in various 3D generation, editing, and inpainting-based tasks.