ComfyUI-FitDiTx

ComfyUI custom nodes based on https://github.com/BoyuanJiang/FitDiT

Installation

cd ComfyUI/custom_nodes
git clone https://github.com/ihmily/ComfyUI-FitDiTx
pip install -r requirements.txt

Download Models

cd ComfyUI/models
pip install huggingface_hub --upgrade
hf download BoyuanJiang/FitDiT --local-dir FitDiT_models

cd ComfyUI/models/clip
hf download openai/clip-vit-large-patch14 --local-dir clip-vit-large-patch14

hf download laion/CLIP-ViT-bigG-14-laion2B-39B-b160k --local-dir CLIP-ViT-bigG-14-laion2B-39B-b160k

---

FitDiT: Advancing the Authentic Garment Details for High-fidelity Virtual Try-on

👋 Join our QQ Chat Group

FitDiT is designed for high-fidelity virtual try-on using Diffusion Transformers (DiT).

Updates

2025/1/16: We provide the ComfyUI version of FitDiT, you can use FitDiT in ComfyUI now.

2025/1/9: We provide a Huggingface Space of FitDiT, thanks for Huggingface community GPU grant for providing the GPU resources.

2024/12/20: The FitDiT model weight is available.

2024/12/17: Inference code is released.

2024/12/4: Our Online Demo is released.

2024/11/25: Our Complex Virtual Dressing Dataset (CVDD) is released.

2024/11/15: Our FitDiT paper is available.

Gradio Demo

Our algorithm is divided into two steps. The first step is to generate the mask of the try-on area, and the second step is to try-on in the mask area.

Step1: Run Mask

You can simpley get try-on mask by click Step1: Run Mask at the right side of gradio demo. If the automatically generated mask are not well covered the area where you want to try-on, you can either adjust the mask by:

Drag the slider of *mask offset top*, *mask offset bottom*, *mask offset left* or *mask offset right* and then click Step1: Run Mask button, this will re-generate mask.

Using the brush or eraser tool to edit the automatically generated mask

Step2: Run Try-on

After generating a suitable mask, you can get the try-on results by click Step2: Run Try-on. In the Try-on resolution drop-down box, you can select a suitable processing resolution. In our online demo, the default resolution is 1152×1536, which means that the input model image and garment image will be pad and resized to this resolution before being fed into the model.

Local Demo

First apply access of FitDiT model weight, then clone model to *local_model_dir*

Enviroment

We test our model with following enviroment

torch==2.4.0
torchvision==0.19.0
diffusers==0.31.0
transformers==4.39.3
gradio==5.8.0
onnxruntime-gpu==1.20.1

Run gradio locally

# Run model with bf16 without any offload, fastest inference and most memory
python gradio_sd3.py --model_path local_model_dir

# Run model with fp16
python gradio_sd3.py --model_path local_model_dir --fp16

# Run model with fp16 and cpu offload, moderate inference and moderate memory
python gradio_sd3.py --model_path local_model_dir --fp16 --offload

# Run model with fp16 and aggressive cpu offload, slowest inference and less memory
python gradio_sd3.py --model_path local_model_dir --fp16 --aggressive_offload

Third-Party Creations

We found there’ve been some 3rd party applications or tutorial based on our FitDiT. Many thanks for their contribution to the community!

If you have any related work that you would like to see displayed, please submit it in the issue.

These projects have not been verified by us. If you have any questions, please seek help from the original project authors.

Tutorial

A tutorial of using the comfyui version of FitDiT, from T8star-Aix at youtube or bilibili

Applications

Local one-click integration package of FitDiT, which can be found at deepface forum

Star History

[](https://star-history.com/#BoyuanJiang/FitDiT&Date)

Contact

This model can only be used for non-commercial use. For commercial use, please visit Tencent Cloud for support.

Citation

If you find our work helpful for your research, please consider citing our work.

@misc{jiang2024fitditadvancingauthenticgarment,
      title={FitDiT: Advancing the Authentic Garment Details for High-fidelity Virtual Try-on}, 
      author={Boyuan Jiang and Xiaobin Hu and Donghao Luo and Qingdong He and Chengming Xu and Jinlong Peng and Jiangning Zhang and Chengjie Wang and Yunsheng Wu and Yanwei Fu},
      year={2024},
      eprint={2411.10499},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2411.10499}, 
}