Skip to content

UniVG

A unified multimodal system for video generation

Visit official site

What UniVG is

The model is described as a unified multimodal video generation system that handles multiple tasks with text and image inputs. The material states multi-condition cross attention and biased Gaussian noise support both high-freedom and low-freedom video generation.

What you can do with it

  • Handle multiple video generation tasks
  • Work with text and image inputs
  • Use multi-condition cross attention and biased Gaussian noise

Model notes

  • High-freedom and low-freedom video generation are supported
  • The material reports the lowest FVD on the public MSR-VTT benchmark
  • Results are compared with open-source methods and the closed-source Gen2 method

Pros & cons

✓ What we like

  • Handles high-freedom and low-freedom video generation
  • Combines text and image modalities in one system

! What to watch out for

  • Generation relies on multi-condition cross attention and biased Gaussian noise
  • The reported FVD result is measured on the MSR-VTT benchmark

FAQ

What modalities does it handle?

The material states text and image modalities.

Which benchmark is mentioned?

The material reports the lowest FVD on MSR-VTT.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review