Icon USO
Unified Style and Subject Driven Generation
 via Disentangled and Reward Learning


Shaojin Wu  Mengqi Huang  Yufeng Cheng  Wenxu Wu  Jiahe Tian  Yiming Luo  Fei Ding  Qian He
UXO Team
Intelligent Creation Lab, ByteDance

Paper arXiv USO SVG Icon Model SVG Icon Dataset SVG Icon Demo UNO

✨ Press and hold the mouse button to drag and view

We announce USO, a unified style-subject optimized customization model and the latest addition to the UXO family. USO can freely combine arbitrary subjects with arbitrary styles in any scenarios, delivering outputs with high subject/identity consistency and strong style fidelity while ensuring natural, non-plastic portraits. In line with our past practice, we will open-source the full project, including training code, inference scripts, model weights, and datasets, to advance research and empower the open-source community.

What can USO do?


Original
Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8 Image 9 Image 10 Image 11 Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8 Image 9 Image 10 Image 11
Original
Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8 Image 9 Image 10 Image 11 Image 12 Image 13 Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8 Image 9 Image 10 Image 11 Image 12 Image 13
Original
Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8 Image 9 Image 10 Image 11 Image 12 Image 13 Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8 Image 9 Image 10 Image 11 Image 12 Image 13
Original
Image 0 Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8 Image 9 Image 10 Image 11 Image 0 Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8 Image 9 Image 10 Image 11
Original
Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8 Image 9 Image 10 Image 11 Image 12 Image 13 Image 14 Image 15 Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8 Image 9 Image 10 Image 11 Image 12 Image 13 Image 14 Image 15
Original
Image 0 Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8 Image 9 Image 10 Image 11 Image 12 Image 13 Image 14 Image 15 Image 16 Image 0 Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8 Image 9 Image 10 Image 11 Image 12 Image 13 Image 14 Image 15 Image 16
Original
Original
Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8 Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8
Original
Original
Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8 Image 1 Image 2 Image 3 Image 4 Image 5 Image 6 Image 7 Image 8

Abstract


Existing literature typically treats style-driven and subject-driven generation as two disjoint tasks: the former prioritizes stylistic similarity, whereas the latter insists on subject consistency, resulting in an apparent antagonism. We argue that both objectives can be unified under a single framework because they ultimately concern the disentanglement and re-composition of content and style, a long-standing theme in style-driven research. To this end, we present USO, a Unified Style-Subject Optimized customization model. First, we construct a large-scale triplet dataset consisting of content images, style images, and their corresponding identity1d content images. Second, we introduce a disentangled learning scheme that simultaneously aligns style features and disentangles content from style through two complementary objectives, style-alignment training and content-style disentanglement training. Third, we incorporate a style reward-learning paradigm denoted as SRL to further enhance the model's performance. Finally, we release USO-Bench, the first benchmark that jointly evaluates style similarity and subject fidelity across multiple metrics. Extensive experiments demonstrate that USO achieves state-of-the-art performance among open-source models along both dimensions of subject consistency and style similarity.


How does it work?



USO unifies subject-driven and style-driven generation in two stages: Stage 1 aligns SigLIP embeddings via style-alignment training to yield a style-capable model; Stage 2 disentangles the conditional encoders and trains on triplets to enable the joint conditional generation. Finally, a style-reward learning paradigm supervises both stages to yield a stronger unified model.



Comparison with State-of-the-Art Methods


Qualitative comparison with different methods on subject-driven generation.


Qualitative comparison with different methods on style-driven generation..


Qualitative comparison with different methods on identity-driven generation


Qualitative comparison with different methods on style-subject-driven generation.

Disclaimer


We open-source this project for academic research. The vast majority of images used in this project are either generated or from open-source datasets. If you have any concerns, please contact us, and we will promptly remove any inappropriate content. Our project is released under the Apache 2.0 License. If you apply to other base models, please ensure that you comply with the original licensing terms.

This research aims to advance the field of generative AI. Users are free to create images using this tool, provided they comply with local laws and exercise responsible usage. The developers are not liable for any misuse of the tool by users.


✨ If you find this project useful for your research, please consider citing our paper: 
@article{wu2025uso,
    title={USO: Unified Style and Subject-Driven Generation via Disentangled and Reward Learning},
    author={Shaojin Wu and Mengqi Huang and Yufeng Cheng and Wenxu Wu and Jiahe Tian and Yiming Luo and Fei Ding and Qian He},
    year={2025},
    eprint={2508.18966},
    archivePrefix={arXiv},
    primaryClass={cs.CV},
}


Source code of the project page can be found in USO Github Pages.