Jing He1✱, Haodong Li1✱, Yongzhe Hu, Guibao Shen1, Yingjie Cai3, Weichao Qiu3, Ying-Cong Chen1,2✉
1HKUST(GZ)
2HKUST
3Noah's Ark Lab
✱Both authors contributed equally.
✉Corresponding author.
We present DisEnvisioner, without cumbersome tuning or relying on multiple reference images, DisEnvisioner is capable of generating a variety of exceptional customized images. Characterized by its emphasis on the interpretation of subject-essential attributes, DisEnvisioner effectively discerns and enhances the subject-essential feature while filtering out irrelevant attributes, achieving superior personalizing quality in both editability and ID consistency.
2024-10-04: Paper released, the code & demo will be available soon!
If you find our work useful in your research, please consider citing our paper:
@article{he2024disenvisioner,
title={DisEnvisioner: Disentangled and Enriched Visual Prompt for Customized Image Generation},
author={Jing He and Haodong Li and Yongzhe Hu and Guibao Shen and Yingjie Cai and Weichao Qiu and Ying-Cong Chen},
journal={arXiv preprint arXiv:2410.02067},
year={2024}
}