DynVideo-E: Harnessing Dynamic NeRF for Large-Scale Motion- and View-Change Human-Centric Video Editing

TL;DR

DynVideo-E introduces dynamic NeRF for large-scale, human-centric video editing, enabling consistent, controllable edits across long videos with significant motion and view changes by leveraging 3D space representations.

Contribution

It pioneers the use of dynamic NeRF with innovative editing pipeline components for long-range, human-centric video editing with superior performance.

Findings

Outperforms state-of-the-art methods by 50-95% in human preference.

Effectively handles large-scale motion and view changes in videos.

Enables consistent, controllable editing in long, complex videos.

Abstract

Despite recent progress in diffusion-based video editing, existing methods are limited to short-length videos due to the contradiction between long-range consistency and frame-wise editing. Prior attempts to address this challenge by introducing video-2D representations encounter significant difficulties with large-scale motion- and view-change videos, especially in human-centric scenarios. To overcome this, we propose to introduce the dynamic Neural Radiance Fields (NeRF) as the innovative video representation, where the editing can be performed in the 3D spaces and propagated to the entire video via the deformation field. To provide consistent and controllable editing, we propose the image-based video-NeRF editing pipeline with a set of innovative designs, including multi-view multi-pose Score Distillation Sampling (SDS) from both the 2D personalized diffusion prior and 3D diffusion prior, reconstruction losses, text-guided local parts super-resolution, and style transfer. Extensive experiments demonstrate that our method, dubbed as DynVideo-E, significantly outperforms SOTA approaches on two challenging datasets by a large margin of 50% ~ 95% for human preference. Code will be released at https://showlab.github.io/DynVideo-E/.