Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DCDM: Divide-and-Conquer Diffusion Models for Consistency-Preserving Video Generation

Haoyu Zhao, Yuang Zhang, Junqi Cheng, Jiaxi Gu, Zenghui Lu, Peng Shu, Zuxuan Wu, Yu-Gang Jiang

TL;DR

This paper proposes a system-level framework, termed the Divide-and-Conquer Diffusion Model (DCDM), to address three key challenges: intra-clip world knowledge consistency, inter-clip camera consistency, and (3) inter-shot element consistency.

Abstract

Recent video generative models have demonstrated impressive visual fidelity, yet they often struggle with semantic, geometric, and identity consistency. In this paper, we propose a system-level framework, termed the Divide-and-Conquer Diffusion Model (DCDM), to address three key challenges: (1) intra-clip world knowledge consistency, (2) inter-clip camera consistency, and (3) inter-shot element consistency. DCDM decomposes video consistency modeling under these scenarios into three dedicated components while sharing a unified video generation backbone. For intra-clip consistency, DCDM leverages a large language model to parse input prompts into structured semantic representations, which are subsequently translated into coherent video content by a diffusion transformer. For inter-clip camera consistency, we propose a temporal camera representation in the noise space that enables precise and stable camera motion control, along with a text-to-image initialization mechanism to further enhance controllability. For inter-shot consistency, DCDM adopts a holistic scene generation paradigm with windowed cross-attention and sparse inter-shot self-attention, ensuring long-range narrative coherence while maintaining computational efficiency. We validate our framework on the test set of the CVM Competition at AAAI'26, and the results demonstrate that the proposed strategies effectively address these challenges.

DCDM: Divide-and-Conquer Diffusion Models for Consistency-Preserving Video Generation

TL;DR

This paper proposes a system-level framework, termed the Divide-and-Conquer Diffusion Model (DCDM), to address three key challenges: intra-clip world knowledge consistency, inter-clip camera consistency, and (3) inter-shot element consistency.

Abstract

Recent video generative models have demonstrated impressive visual fidelity, yet they often struggle with semantic, geometric, and identity consistency. In this paper, we propose a system-level framework, termed the Divide-and-Conquer Diffusion Model (DCDM), to address three key challenges: (1) intra-clip world knowledge consistency, (2) inter-clip camera consistency, and (3) inter-shot element consistency. DCDM decomposes video consistency modeling under these scenarios into three dedicated components while sharing a unified video generation backbone. For intra-clip consistency, DCDM leverages a large language model to parse input prompts into structured semantic representations, which are subsequently translated into coherent video content by a diffusion transformer. For inter-clip camera consistency, we propose a temporal camera representation in the noise space that enables precise and stable camera motion control, along with a text-to-image initialization mechanism to further enhance controllability. For inter-shot consistency, DCDM adopts a holistic scene generation paradigm with windowed cross-attention and sparse inter-shot self-attention, ensuring long-range narrative coherence while maintaining computational efficiency. We validate our framework on the test set of the CVM Competition at AAAI'26, and the results demonstrate that the proposed strategies effectively address these challenges.
Paper Structure (20 sections, 9 equations, 2 figures)

This paper contains 20 sections, 9 equations, 2 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Method
  4. Preliminaries
  5. Intra-clip World Knowledge Consistency
  6. Motivation.
  7. LLM-based Prompt Extension.
  8. Inter-clip Camera Consistency via Camera Control
  9. Motivation.
  10. Camera Representation Construction
  11. Noise-space Injection
  12. Reference Image Conditioning
  13. Inter-shot Element Consistency with Sparse Inter-shot Self-Attention
  14. Motivation.
  15. Sparse Inter-shot Self-Attention.
  16. ...and 5 more sections

Figures (2)

  • Figure 1: A high-level overview of our proposed Divide-and-Conquer Diffusion Model (DCDM) framework.
  • Figure 2: Qualitative results of our DCDM framework across intra-clip world knowledge consistency (above), inter-clip camera consistency (middle), and inter-shot element consistency scenarios (bottom).