Skip to content

ViViD

A diffusion framework for video virtual try-on

Visit official site

What ViViD is

The framework is described as a new video virtual try-on approach built on diffusion models, using a garment encoder for fine semantic features and a lightweight pose encoder for spatiotemporal consistency. Its dataset is described as large-scale, with diverse garment types and high resolution.

What you can do with it

  • Extract fine garment semantic features with a garment encoder
  • Inject garment detail into the target video with attention feature fusion
  • Encode pose signals to learn garment and body pose interaction
  • Generate coherent video with a temporal module in a text-to-image diffusion model

Getting started

  • Visit the project page and download the code and dataset
  • Install the required dependencies and environment from the docs
  • Run the garment encoder to extract garment features
  • Process the target video with the pose encoder for body pose data

Application scenarios

  • An online retailer offers personalised virtual try-on
  • A clothing designer shows new designs to potential buyers
  • A video creator adds interactive and playful try-on content

Pros & cons

✓ What we like

  • The material states code, dataset and weights will be made public
  • A large dataset with diverse garment types is described

! What to watch out for

  • The code, dataset and weights are listed as forthcoming releases
  • Video output depends on the garment and pose encoders

FAQ

What does the framework generate?

The material describes realistic video virtual try-on results.

What will be released?

The material states the code, dataset and weights will be made publicly available.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review