PuLID: Pure and Lightning ID Customization via Contrastive Alignment
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866917824287473664 |
|---|---|
| author | Guo, Zinan Wu, Yanze Chen, Zhuowei Chen, Lang Zhang, Peng He, Qian |
| author_facet | Guo, Zinan Wu, Yanze Chen, Zhuowei Chen, Lang Zhang, Peng He, Qian |
| contents | We propose Pure and Lightning ID customization (PuLID), a novel tuning-free ID customization method for text-to-image generation. By incorporating a Lightning T2I branch with a standard diffusion one, PuLID introduces both contrastive alignment loss and accurate ID loss, minimizing disruption to the original model and ensuring high ID fidelity. Experiments show that PuLID achieves superior performance in both ID fidelity and editability. Another attractive property of PuLID is that the image elements (e.g., background, lighting, composition, and style) before and after the ID insertion are kept as consistent as possible. Codes and models are available at https://github.com/ToTheBeginning/PuLID |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2404_16022 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | PuLID: Pure and Lightning ID Customization via Contrastive Alignment Guo, Zinan Wu, Yanze Chen, Zhuowei Chen, Lang Zhang, Peng He, Qian Computer Vision and Pattern Recognition We propose Pure and Lightning ID customization (PuLID), a novel tuning-free ID customization method for text-to-image generation. By incorporating a Lightning T2I branch with a standard diffusion one, PuLID introduces both contrastive alignment loss and accurate ID loss, minimizing disruption to the original model and ensuring high ID fidelity. Experiments show that PuLID achieves superior performance in both ID fidelity and editability. Another attractive property of PuLID is that the image elements (e.g., background, lighting, composition, and style) before and after the ID insertion are kept as consistent as possible. Codes and models are available at https://github.com/ToTheBeginning/PuLID |
| title | PuLID: Pure and Lightning ID Customization via Contrastive Alignment |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2404.16022 |