Skip to content

IDM-VTON

Generate virtual try-on images from a person and a garment

Visit official site

What IDM-VTON is

The model is described as a diffusion model for image-based virtual try-on that joins high-level semantics from a visual encoder with low-level features from a parallel UNet. The material states that detailed text prompts and a customisation method lift the fidelity and realism of try-on images in real-world scenes.

What you can do with it

  • Generate a virtual try-on image from a person and a garment image
  • Encode garment image semantics with two separate modules
  • Fuse high-level semantics through cross-attention layers
  • Fuse low-level features from a parallel UNet through self-attention layers

Application scenarios

  • A clothing brand shows a new season collection on different models
  • An online shop offers customers a personal virtual try-on service
  • A fashion blogger shares outfit suggestions on social media
  • A clothing designer shows garments on different people

Getting started

  • Prepare a person image and a garment image
  • Provide detailed text prompts for the garment and the person
  • Generate the virtual try-on image with the model
  • Refine the result with the customisation method for a real-world scene

Pros & cons

✓ What we like

  • Detailed text prompts are used to improve realism
  • A customisation method targets fidelity in real-world scenes

! What to watch out for

  • The workflow starts from a pair of person and garment images
  • Output fidelity relies on the customisation method

FAQ

What does the model take as input?

A person image, a garment image and detailed text prompts for both.

What is the customisation method for?

The material says it improves fidelity and realism in real-world scenes.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review