ComposeMe: Attribute-Specific Image Prompts for Controllable Human Image Generation
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866917017908412416 |
|---|---|
| author | Qian, Guocheng Gordon Ostashev, Daniil Nemchinov, Egor Assouline, Avihay Tulyakov, Sergey Wang, Kuan-Chieh Jackson Aberman, Kfir |
| author_facet | Qian, Guocheng Gordon Ostashev, Daniil Nemchinov, Egor Assouline, Avihay Tulyakov, Sergey Wang, Kuan-Chieh Jackson Aberman, Kfir |
| contents | Generating high-fidelity images of humans with fine-grained control over attributes such as hairstyle and clothing remains a core challenge in personalized text-to-image synthesis. While prior methods emphasize identity preservation from a reference image, they lack modularity and fail to provide disentangled control over specific visual attributes. We introduce a new paradigm for attribute-specific image prompting, in which distinct sets of reference images are used to guide the generation of individual aspects of human appearance, such as hair, clothing, and identity. Our method encodes these inputs into attribute-specific tokens, which are injected into a pre-trained text-to-image diffusion model. This enables compositional and disentangled control over multiple visual factors, even across multiple people within a single image. To promote natural composition and robust disentanglement, we curate a cross-reference training dataset featuring subjects in diverse poses and expressions, and propose a multi-attribute cross-reference training strategy that encourages the model to generate faithful outputs from misaligned attribute inputs while adhering to both identity and textual conditioning. Extensive experiments show that our method achieves state-of-the-art performance in accurately following both visual and textual prompts. Our framework paves the way for more configurable human image synthesis by combining visual prompting with text-driven generation. Webpage is available at: https://snap-research.github.io/composeme/. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_18092 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | ComposeMe: Attribute-Specific Image Prompts for Controllable Human Image Generation Qian, Guocheng Gordon Ostashev, Daniil Nemchinov, Egor Assouline, Avihay Tulyakov, Sergey Wang, Kuan-Chieh Jackson Aberman, Kfir Computer Vision and Pattern Recognition Generating high-fidelity images of humans with fine-grained control over attributes such as hairstyle and clothing remains a core challenge in personalized text-to-image synthesis. While prior methods emphasize identity preservation from a reference image, they lack modularity and fail to provide disentangled control over specific visual attributes. We introduce a new paradigm for attribute-specific image prompting, in which distinct sets of reference images are used to guide the generation of individual aspects of human appearance, such as hair, clothing, and identity. Our method encodes these inputs into attribute-specific tokens, which are injected into a pre-trained text-to-image diffusion model. This enables compositional and disentangled control over multiple visual factors, even across multiple people within a single image. To promote natural composition and robust disentanglement, we curate a cross-reference training dataset featuring subjects in diverse poses and expressions, and propose a multi-attribute cross-reference training strategy that encourages the model to generate faithful outputs from misaligned attribute inputs while adhering to both identity and textual conditioning. Extensive experiments show that our method achieves state-of-the-art performance in accurately following both visual and textual prompts. Our framework paves the way for more configurable human image synthesis by combining visual prompting with text-driven generation. Webpage is available at: https://snap-research.github.io/composeme/. |
| title | ComposeMe: Attribute-Specific Image Prompts for Controllable Human Image Generation |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2509.18092 |