Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2506.21416 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909660728000512 |
|---|---|
| author | Chen, Bowen Zhao, Mengyi Sun, Haomiao Chen, Li Wang, Xu Du, Kang Wu, Xinglong |
| author_facet | Chen, Bowen Zhao, Mengyi Sun, Haomiao Chen, Li Wang, Xu Du, Kang Wu, Xinglong |
| contents | Achieving fine-grained control over subject identity and semantic attributes (pose, style, lighting) in text-to-image generation, particularly for multiple subjects, often undermines the editability and coherence of Diffusion Transformers (DiTs). Many approaches introduce artifacts or suffer from attribute entanglement. To overcome these challenges, we propose a novel multi-subject controlled generation model XVerse. By transforming reference images into offsets for token-specific text-stream modulation, XVerse allows for precise and independent control for specific subject without disrupting image latents or features. Consequently, XVerse offers high-fidelity, editable multi-subject image synthesis with robust control over individual subject characteristics and semantic attributes. This advancement significantly improves personalized and complex scene generation capabilities. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_21416 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | XVerse: Consistent Multi-Subject Control of Identity and Semantic Attributes via DiT Modulation Chen, Bowen Zhao, Mengyi Sun, Haomiao Chen, Li Wang, Xu Du, Kang Wu, Xinglong Computer Vision and Pattern Recognition Achieving fine-grained control over subject identity and semantic attributes (pose, style, lighting) in text-to-image generation, particularly for multiple subjects, often undermines the editability and coherence of Diffusion Transformers (DiTs). Many approaches introduce artifacts or suffer from attribute entanglement. To overcome these challenges, we propose a novel multi-subject controlled generation model XVerse. By transforming reference images into offsets for token-specific text-stream modulation, XVerse allows for precise and independent control for specific subject without disrupting image latents or features. Consequently, XVerse offers high-fidelity, editable multi-subject image synthesis with robust control over individual subject characteristics and semantic attributes. This advancement significantly improves personalized and complex scene generation capabilities. |
| title | XVerse: Consistent Multi-Subject Control of Identity and Semantic Attributes via DiT Modulation |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2506.21416 |