Lumina-Video: Efficient and Flexible Video Generation with Multi-scale Next-DiT

TL;DR

Lumina-Video introduces a multi-scale Next-DiT framework for efficient, flexible, and controllable high-quality video generation, leveraging tailored architectures and training schemes to address spatiotemporal complexity.

Contribution

It presents Lumina-Video with a multi-scale Next-DiT architecture, motion control, progressive training, and a novel video-to-audio model, advancing video synthesis capabilities.

Findings

High aesthetic quality and motion smoothness achieved

Efficient training and inference at high resolutions and FPS

Effective control of generated video dynamics

Abstract

Recent advancements have established Diffusion Transformers (DiTs) as a dominant framework in generative modeling. Building on this success, Lumina-Next achieves exceptional performance in the generation of photorealistic images with Next-DiT. However, its potential for video generation remains largely untapped, with significant challenges in modeling the spatiotemporal complexity inherent to video data. To address this, we introduce Lumina-Video, a framework that leverages the strengths of Next-DiT while introducing tailored solutions for video synthesis. Lumina-Video incorporates a Multi-scale Next-DiT architecture, which jointly learns multiple patchifications to enhance both efficiency and flexibility. By incorporating the motion score as an explicit condition, Lumina-Video also enables direct control of generated videos' dynamic degree. Combined with a progressive training scheme with increasingly higher resolution and FPS, and a multi-source training scheme with mixed natural and synthetic data, Lumina-Video achieves remarkable aesthetic quality and motion smoothness at high training and inference efficiency. We additionally propose Lumina-V2A, a video-to-audio model based on Next-DiT, to create synchronized sounds for generated videos. Codes are released at https://www.github.com/Alpha-VLLM/Lumina-Video.