GeoNVS: Geometry Grounded Video Diffusion for Novel View Synthesis

TL;DR

GeoNVS introduces a geometry-grounded video diffusion method that significantly improves novel view synthesis by explicitly incorporating 3D geometric guidance, leading to better geometric fidelity and camera control.

Contribution

The paper proposes GS-Adapter, a novel feature adapter that lifts 2D features into 3D Gaussian representations for improved view synthesis without additional training.

Findings

Achieves state-of-the-art performance across multiple scenes and settings.

Improves geometric accuracy with up to 2x reduction in translation error.

Outperforms prior methods by 11.3% and 14.9% in key metrics.

Abstract

Novel view synthesis requires strong 3D geometric consistency and the ability to generate visually coherent images across diverse viewpoints. While recent camera-controlled video diffusion models show promising results, they often suffer from geometric distortions and limited camera controllability. To overcome these challenges, we introduce GeoNVS, a geometry-grounded novel-view synthesizer that enhances both geometric fidelity and camera controllability through explicit 3D geometric guidance. Our key innovation is the Gaussian Splat Feature Adapter (GS-Adapter), which lifts input-view diffusion features into 3D Gaussian representations, renders geometry-constrained novel-view features, and adaptively fuses them with diffusion features to correct geometrically inconsistent representations. Unlike prior methods that inject geometry at the input level, GS-Adapter operates in feature space, avoiding view-dependent color noise that degrades structural consistency. Its plug-and-play design enables zero-shot compatibility with diverse feed-forward geometry models without additional training, and can be adapted to other video diffusion backbones. Experiments across 9 scenes and 18 settings demonstrate state-of-the-art performance, achieving 11.3% and 14.9% improvements over SEVA and CameraCtrl, with up to 2x reduction in translation error and 7x in Chamfer Distance.