V-RGBX: Video Editing with Accurate Controls over Intrinsic Properties

TL;DR

V-RGBX is an innovative end-to-end framework that enables intrinsic-aware video editing by jointly understanding, synthesizing, and editing intrinsic scene properties for photorealistic and temporally consistent results.

Contribution

It introduces the first unified framework for intrinsic-aware video editing, combining inverse rendering, synthesis, and keyframe-based editing conditioned on intrinsic properties.

Findings

Produces temporally consistent, photorealistic videos.

Supports flexible manipulation of intrinsic modalities.

Outperforms prior methods in object editing and relighting.

Abstract

Large-scale video generation models have shown remarkable potential in modeling photorealistic appearance and lighting interactions in real-world scenes. However, a closed-loop framework that jointly understands intrinsic scene properties (e.g., albedo, normal, material, and irradiance), leverages them for video synthesis, and supports editable intrinsic representations remains unexplored. We present V-RGBX, the first end-to-end framework for intrinsic-aware video editing. V-RGBX unifies three key capabilities: (1) video inverse rendering into intrinsic channels, (2) photorealistic video synthesis from these intrinsic representations, and (3) keyframe-based video editing conditioned on intrinsic channels. At the core of V-RGBX is an interleaved conditioning mechanism that enables intuitive, physically grounded video editing through user-selected keyframes, supporting flexible manipulation of any intrinsic modality. Extensive qualitative and quantitative results show that V-RGBX produces temporally consistent, photorealistic videos while propagating keyframe edits across sequences in a physically plausible manner. We demonstrate its effectiveness in diverse applications, including object appearance editing and scene-level relighting, surpassing the performance of prior methods.