SwiftPie: Lightning-fast Subject-driven Image Personalization via One step Diffusion
Huy Duong Trong-Tung Nguyen Cuong Pham Anh Tran Khoi Nguyen Minh Hoai · 2026
Read SwiftPie: Lightning-fast Subject-driven Image Personalization via One step Diffusion by Huy Duong Trong-Tung Nguyen Cuong Pham Anh Tran Khoi Nguyen Minh Hoai (2026) 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-2605-01510.txt.
Description
Huy Duong and five co-authors make subject-driven image generation run in a single step. Given a photo of a person or pet, personalization tools redraw that subject in new scenes, but they lean on slow fine-tuning, long optimization, or many denoising steps that block real-time use.
SwiftPie is presented as the first one-step diffusion tool for the job. A dual-branch identity injection feeds the subject's look into the single pass, and a mask-guided rescaling step strengthens how the subject sits in its new context. Across tests it runs far faster than multi-step rivals while matching them on keeping the subject's identity and following the prompt. The authors frame the speed as the missing piece for interactive tools, where a user can steer a personal image as fast as they can type.
Catalog record
| CREATOR | Huy Duong Trong-Tung Nguyen Cuong Pham Anh Tran Khoi Nguyen Minh Hoai |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 39 KB |
| LICENSE | http://creativecommons.org/licenses/by-sa/4.0/ |
| RIGHTS | CC BY SA |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.01510