Versatile Editing of Video Content, Actions, and Dynamics without Training

TL;DR

DynaEdit is a training-free, versatile video editing method that leverages pretrained text-to-video flow models to modify actions, insert objects, and apply effects without retraining, overcoming previous limitations.

Contribution

It introduces DynaEdit, a novel training-free approach using inversion-free models for complex, unconstrained video editing tasks with state-of-the-art results.

Findings

Achieves high-quality editing of actions and interactions in videos.

Handles complex modifications like object insertion and global effects.

Outperforms existing methods on benchmark tasks.

Abstract

Controlled video generation has seen drastic improvements in recent years. However, editing actions and dynamic events, or inserting contents that should affect the behaviors of other objects in real-world videos, remains a major challenge. Existing trained models struggle with complex edits, likely due to the difficulty of collecting relevant training data. Similarly, existing training-free methods are inherently restricted to structure- and motion-preserving edits and do not support modification of motion or interactions. Here, we introduce DynaEdit, a training-free editing method that unlocks versatile video editing capabilities with pretrained text-to-video flow models. Our method relies on the recently introduced inversion-free approach, which does not intervene in the model internals, and is thus model-agnostic. We show that naively attempting to adapt this approach to general unconstrained editing results in severe low-frequency misalignment and high-frequency jitter. We explain the sources for these phenomena and introduce novel mechanisms for overcoming them. Through extensive experiments, we show that DynaEdit achieves state-of-the-art results on complex text-based video editing tasks, including modifying actions, inserting objects that interact with the scene, and introducing global effects.