Cozy Clothes Swap - Customizable ComfyUI Node For Fashion Try-on With SAL-VTON

A ComfyUI node to dress-up your friends or your characters. Based on a warping algorithm.

Made with 💚 by the CozyMantis squad.

Installation

• Copy all the node files into the custom_nodes/cozy-sal-vton directory, then run pip install -r requirements.txt to install the required dependencies.
• Download the required models from the official release by using git: git clone https://www.modelscope.cn/iic/cv_SAL-VTON_virtual-try-on.git
• Copy all the files in the above repository to the models/sal-vton directory.

Usage

An example workflow is available here.

The SAL-VTON models have been trained on the VITON-HD dataset, so for best results you'll want:

• images that have a white/light gray background
• upper-body clothing items (tops, tshirts, bodysuits, etc.)
• an input person standing up straight, pictured from the knees/thighs up.

The virtual try-on node takes three images as input:

• the first image is the person you want to dress up
• the second image is the clothing item you want to put on them
• the third image is the mask of the clothing item (generate this via any background removal node)

There are two stages to the process:

• in the first stage, coordinates are generated for a series of predefined landmarks on the person's body and the clothing item
• in the second stage, the landmark coordinates are used to warp the garment and generate a new image of the person wearing the clothing item

The node will output:

• the image of the person wearing the clothing item
• an image displaying the landmark coordinates for the person
• an image displaying the landmark coordinates for the clothing item

The coordinates will be auto-generated the first time you run the node. If needed, you can correct the fit by manually adjusting the landmark coordinates and re-running the workflow. Press the "Update Landmarks" button to bring up the landmark editor.

Press one of the buttons in the left column to select a specific landmark, then click on the images to assign it.

Each landmark also needs to be assigned a category for each input. The categories are represented in the dropdowns next to each landmark:

• the first dropdown represents the person landmark category, and is one of: 0 (if the landmark does not exist in the provided image), 1 (if the landmark exists but is occluded by other parts of the body), or 2 (if the landmark exists and is not occluded)
• the second dropdown represents the clothing item landmark category, and is one of: 0 (if the landmark does not exist in the provided image), or 1 (if the landmark exists)

Important

Make sure you own the rights to the images you use with this node. Do not use images that you do not have permission to use.

Acknowledgements

Based on the excellent paper "Linking Garment With Person via Semantically Associated Landmarks for Virtual Try-On" by Keyu Yan, Tingwei Gao, HUI ZHANG, Chengjun Xie.