Uni-ViGU: Towards Unified Video Generation and Understanding via A Diffusion-Based Video Generator

TL;DR

Uni-ViGU introduces a unified framework for video generation and understanding that leverages a diffusion-based generator and a novel training mechanism to achieve competitive multimodal performance.

Contribution

It proposes a new unified approach extending a video generator with a flow matching method and a modality-driven MoE framework for joint video understanding and generation.

Findings

Achieves competitive results in video generation and understanding tasks.

Introduces a bidirectional training mechanism for shared representations.

Demonstrates the scalability of generation-centric architectures for multimodal intelligence.

Abstract

Unified multimodal models integrating visual understanding and generation face a fundamental challenge: visual generation incurs substantially higher computational costs than understanding, particularly for video. This imbalance motivates us to invert the conventional paradigm: rather than extending understanding-centric MLLMs to support generation, we propose Uni-ViGU, a framework that unifies video generation and understanding by extending a video generator as the foundation. We introduce a unified flow method that performs continuous flow matching for video and discrete flow matching for text within a single process, enabling coherent multimodal generation. We further propose a modality-driven MoE-based framework that augments Transformer blocks with lightweight layers for text generation while preserving generative priors. To repurpose generation knowledge for understanding, we design a bidirectional training mechanism with two stages: Knowledge Recall reconstructs input prompts to leverage learned text-video correspondences, while Capability Refinement fine-tunes on detailed captions to establish discriminative shared representations. Experiments demonstrate that Uni-ViGU achieves competitive performance on both video generation and understanding, validating generation-centric architectures as a scalable path toward unified multimodal intelligence. Project Page and Code: https://fr0zencrane.github.io/uni-vigu-page/.