AIComposer: Any Style and Content Image Composition via Feature Integration (ICCV 2025)

Installation

# install diffusers
pip install diffusers==0.29.2

# download the models
cd AIComposer
git lfs install
git clone https://huggingface.co/sherlhw/AIComposer
mv AIComposer mlp_model

Download Models

• h94/IP-Adapter
• stabilityai/sd-vae-ft-mse
• stabilityai/stable-diffusion-xl-base-1.0

Running AIComposer

Data Preparation

Several input samples are available under ./examples directory. Each sample involves one background (background_image), one foreground (ref_image), one segmentation mask for the foreground (ref_mask), and one user mask that denotes the desired composition location (background_mask).

More samples are available in AIComposer Test Benchmark or you can customize them.

Image Composition

python demo.py # SDXL版本

Acknowledgments

Our work is standing on the shoulders of giants. We thank the following contributors that our code is based on: diffusers, IP-Adapter, TF-ICON and Magic Insert.

Citation

If you find AIComposer useful for your research and applications, please cite using this BibTeX:

@article{li2025aicomposerstylecontentimage,
      title={AIComposer: Any Style and Content Image Composition via Feature Integration}, 
      author={Haowen Li and Zhenfeng Fan and Zhang Wen and Zhengzhou Zhu and Yunjin Li},
      booktitle={arXiv preprint arxiv:2507.20721},
      year={2025}
}