AutoVFX: Physically Realistic Video Editing from Natural Language Instructions

TL;DR

AutoVFX is a novel framework that enables realistic, physically grounded video editing from a single video using natural language instructions, making VFX creation more accessible and versatile.

Contribution

The paper introduces AutoVFX, integrating neural scene modeling, LLM-based code generation, and physical simulation for natural language-controlled video editing.

Findings

Outperforms competing methods in generative quality

Achieves high instruction alignment and editing versatility

Ensures physical plausibility in edited videos

Abstract

Modern visual effects (VFX) software has made it possible for skilled artists to create imagery of virtually anything. However, the creation process remains laborious, complex, and largely inaccessible to everyday users. In this work, we present AutoVFX, a framework that automatically creates realistic and dynamic VFX videos from a single video and natural language instructions. By carefully integrating neural scene modeling, LLM-based code generation, and physical simulation, AutoVFX is able to provide physically-grounded, photorealistic editing effects that can be controlled directly using natural language instructions. We conduct extensive experiments to validate AutoVFX's efficacy across a diverse spectrum of videos and instructions. Quantitative and qualitative results suggest that AutoVFX outperforms all competing methods by a large margin in generative quality, instruction alignment, editing versatility, and physical plausibility.