Conceptrol: Concept Control of Zero-shot Personalized Image Generation
Qiyuan He, Angela Yao · 2025
Read Conceptrol: Concept Control of Zero-shot Personalized Image Generation by Qiyuan He, Angela Yao (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-06568.txt.
Description
Qiyuan He and Angela Yao present Conceptrol, a fix for image generators that personalize pictures from a reference photo. Their paper appeared in 2025. Add-on modules can insert a person's face or a pet into new scenes without retraining. Yet the output often copies the reference and ignores the text prompt. The authors trace this to weak mixing of the reference image with the prompt inside the module.
Conceptrol adds a mask drawn from the text, which steers where the module may look in the reference. The change adds no extra computing cost. On standard tests, scores rose by as much as 89 percent over the plain add-on. The method even beat approaches that fine-tune the model for each subject. Code is shared openly for other groups to use.
Catalog record
| CREATOR | Qiyuan He, Angela Yao |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 48 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-06568