Existing text-to-image customization methods (i.e., subject-driven generation) face a fundamental challenge due to the entangled influence of visual and textual conditions. This inherent conflict forces a trade-off between subject fidelity and textual controllability, preventing simultaneous optimization of both objectives.We present RealCustom to disentangle subject similarity from text controllability and thereby allows both to be optimized simultaneously without conflicts. The core idea of RealCustom is to represent given subjects as real words that can be seamlessly integrated with given texts, and further leveraging the relevance between real words and image regions to disentangle visual condition from text condition.
- [10/2025] 🔥 RealCustom++ is accepted by T-PAMI.
- [04/2025] 🔥 We release our newly customization framework UNO
- [04/2025] 🔥 The code and model of RealCustom is released.
Install the requirements
bash envs/init.shYou can dowload all the models in huggingface and put them in ckpts/.
single image inference
bash inference/inference_single_image.shbatch image inference
bash inference/inference_batch_images.shpython inference/app.py
our real-word paradigm naturally supports multi-round generation, where the output from each round serves as the reference subject image for the next. This enables flexible customization in each round by specifying different target real words. For example, in the first row, the initial round uses "dog" as the target word, preserving only the dog's characteristics. In the second round, the target word "dog with the pink hat" incorporates the pink hat generated in the previous round, allowing RealCustom++ to retain both features. This demonstrates the strong generalization capability of RealCustom++, enabling the progressive accumulation and preservation of subject characteristics across multiple rounds.
🎨 Enjoy on Dreamina
RealCustom is previously commercially applied in Dreamina and Doubao, ByteDance. You can also enjoy the more advanced customization algorithm in Dreamina!
Create character images and corresponding appearance descriptions through prompt descriptions, uploading reference images. Specifically: 1. Character Image: Best in clean background, close-up, prominent subject, high-quality resolution. 2. Character Description: Brief, includes the subject and key appearance elements.
Input prompts where the subject is replaced by the selected character, guiding the character to make corresponding changes such as style, actions, expressions, scenes, and modifiers. There is no need to add descriptions of the subject in the prompt. "Face Reference Strength" is the weight for ID retention, and "Body Reference Strength" is the weight for IP retention.
If you find this project useful for your research, please consider citing our papers:
@inproceedings{huang2024realcustom,
title={RealCustom: narrowing real text word for real-time open-domain text-to-image customization},
author={Huang, Mengqi and Mao, Zhendong and Liu, Mingcong and He, Qian and Zhang, Yongdong},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
pages={7476--7485},
year={2024}
}
@ARTICLE{11206511,
author={Mao, Zhendong and Huang, Mengqi and Ding, Fei and Liu, Mingcong and He, Qian and Zhang, Yongdong},
journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
title={RealCustom++: Representing Images as Real Textual Word for Real-Time Customization},
year={2025},
volume={},
number={},
pages={1-18},
doi={10.1109/TPAMI.2025.3623025}
}
@article{wu2025less,
title={Less-to-More Generalization: Unlocking More Controllability by In-Context Generation},
author={Wu, Shaojin and Huang, Mengqi and Wu, Wenxu and Cheng, Yufeng and Ding, Fei and He, Qian},
journal={arXiv preprint arXiv:2504.02160},
year={2025}
}