1.The workflow integrates Fluxtools Redux, Fill, and CatvitonLora

(download links for the catvitonlora are provided in the notes)

2. Using this workflow requires basic knowledge of FluxRedux and Fill model configurations.  

3. Virtual try-on can be achieved even if the character's pose and clothing are different.

Tips:

Using the working methods of Redux Advanced and the Add Mask For IC Lora node, you can also choose to use IC Lora as a replacement for CatVitOn Lora, but this will produce different results.

Setting StyleModelApplySimple to the Highest level of image control helps preserve a significant amount of clothing detail at the expense of image quality, necessitating Upscale to enhance image quality. However, during the upscaling process, some details may be lost. Lowering the Denoise value can help retain more details.

The range of mask detection can focus on Clothes or Clothes and Body. If the background is blank, you can also include the Background in the mask. However, selecting a range that is too large may not necessarily preserve the clothing shapes better.

The Restore group is used to restore parts of the image before upscaling, helping to retain the details of the face and background.

==

This workflow with a complete environment package that includes both Chinese and English interface options. To use it on Windows, click run.bat under the environment package.

https://huggingface.co/dseditor/FLUXFillGGUF/resolve/main/ComfyUI_FLUXTryon.zip