The Dawn of Video Generation: Preliminary Explorations with SORA-like Models

TL;DR

This paper explores the advancements and limitations of SORA-like models in high-quality video generation, emphasizing architectural evolution, data strategies, and the need for better evaluation metrics.

Contribution

It provides a comprehensive investigation into SORA-like models' capabilities, limitations, and the challenges in benchmarking and evaluation.

Findings

Models like SORA achieve higher resolution and natural motion.

Architectural shifts from UNet to DiT improve scalability.

Current benchmarks and metrics inadequately reflect human preferences.

Abstract

High-quality video generation, encompassing text-to-video (T2V), image-to-video (I2V), and video-to-video (V2V) generation, holds considerable significance in content creation to benefit anyone express their inherent creativity in new ways and world simulation to modeling and understanding the world. Models like SORA have advanced generating videos with higher resolution, more natural motion, better vision-language alignment, and increased controllability, particularly for long video sequences. These improvements have been driven by the evolution of model architectures, shifting from UNet to more scalable and parameter-rich DiT models, along with large-scale data expansion and refined training strategies. However, despite the emergence of DiT-based closed-source and open-source models, a comprehensive investigation into their capabilities and limitations remains lacking. Furthermore, the rapid development has made it challenging for recent benchmarks to fully cover SORA-like models and recognize their significant advancements. Additionally, evaluation metrics often fail to align with human preferences.