PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
March 18, 20241 citationsOpen Access

VideoMV: Consistent Multi-View Generation Based on Large Video Generative Model

View Full Paper
QZQi ZuoXGXiaodong GuLQLingteng Qiu

Key Points

  • Fine-tuning video generative models produces consistent 24-view outputs while reducing training time to 4 GPU hours compared to several thousand hours for baseline models.
  • Framework uses 3D-aware denoising sampling with feed-forward reconstruction to integrate rendered global 3D model views into the iterative image diffusion generation loop.
  • Fast reconstruction yields 3D Gaussians in seconds and improves multi-view consistency, providing a rapid asset generation pipeline for dense 3D content creation workflows.

Abstract

Generating multi-view images based on text or single-image prompts is a critical capability for the creation of 3D content. Two fundamental questions on this topic are what data we use for training and how to ensure multi-view consistency. This paper introduces a novel framework that makes fundamental contributions to both questions. Unlike leveraging images from 2D diffusion models for training, we propose a dense consistent multi-view generation model that is fine-tuned from off-the-shelf video generative models. Images from video generative models are more suitable for multi-view generation because the underlying network architecture that generates them employs a temporal module to enforce frame consistency. Moreover, the video data sets used to train these models are abundant and diverse, leading to a reduced train-finetuning domain gap. To enhance multi-view consistency, we introduce a 3D-Aware Denoising Sampling, which first employs a feed-forward reconstruction module to get an explicit global 3D model, and then adopts a sampling strategy that effectively involves images rendered from the global 3D model into the denoising sampling loop to improve the multi-view consistency of the final images. As a by-product, this module also provides a fast way to create 3D assets represented by 3D Gaussians within a few seconds. Our approach can generate 24 dense views and converges much faster in training than state-of-the-art approaches (4 GPU hours versus many thousand GPU hours) with comparable visual quality and consistency. By further fine-tuning, our approach outperforms existing state-of-the-art methods in both quantitative metrics and visual effects. Our project page is aigc3d.github.io/VideoMV.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Zuo et al. (2024) studied this question.

synapsesocial.com/papers/68e73a7cb6db6435876b3af4https://doi.org/10.48550/arxiv.2403.12010
Ask AI
Helpful
Bookmark
Share
View Full Paper

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Vivid-ZOO: Multi-View Video Generation with Diffusion Model2024 · 1 citations
  2. 2MVD-Fusion: Single-view 3D via Depth-consistent Multi-view Generation2024
  3. 3MvDeDiffusion: Multi-view Consistent Generation via Cross-view Deformable Attention for Denoising Diffusion Models2025
  4. 4FVGen: Accelerating Novel-View Synthesis with Adversarial Video Diffusion Distillation2025
  5. 5Animate3D: Animating Any 3D Model with Multi-view Video Diffusion2024