DreamStyle: A Unified Framework for Video Stylization
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866912804844339200 |
|---|---|
| author | Li, Mengtian Chen, Jinshu Zhao, Songtao Feng, Wanquan Tu, Pengqi He, Qian |
| author_facet | Li, Mengtian Chen, Jinshu Zhao, Songtao Feng, Wanquan Tu, Pengqi He, Qian |
| contents | Video stylization, an important downstream task of video generation models, has not yet been thoroughly explored. Its input style conditions typically include text, style image, and stylized first frame. Each condition has a characteristic advantage: text is more flexible, style image provides a more accurate visual anchor, and stylized first frame makes long-video stylization feasible. However, existing methods are largely confined to a single type of style condition, which limits their scope of application. Additionally, their lack of high-quality datasets leads to style inconsistency and temporal flicker. To address these limitations, we introduce DreamStyle, a unified framework for video stylization, supporting (1) text-guided, (2) style-image-guided, and (3) first-frame-guided video stylization, accompanied by a well-designed data curation pipeline to acquire high-quality paired video data. DreamStyle is built on a vanilla Image-to-Video (I2V) model and trained using a Low-Rank Adaptation (LoRA) with token-specific up matrices that reduces the confusion among different condition tokens. Both qualitative and quantitative evaluations demonstrate that DreamStyle is competent in all three video stylization tasks, and outperforms the competitors in style consistency and video quality. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_02785 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | DreamStyle: A Unified Framework for Video Stylization Li, Mengtian Chen, Jinshu Zhao, Songtao Feng, Wanquan Tu, Pengqi He, Qian Computer Vision and Pattern Recognition Video stylization, an important downstream task of video generation models, has not yet been thoroughly explored. Its input style conditions typically include text, style image, and stylized first frame. Each condition has a characteristic advantage: text is more flexible, style image provides a more accurate visual anchor, and stylized first frame makes long-video stylization feasible. However, existing methods are largely confined to a single type of style condition, which limits their scope of application. Additionally, their lack of high-quality datasets leads to style inconsistency and temporal flicker. To address these limitations, we introduce DreamStyle, a unified framework for video stylization, supporting (1) text-guided, (2) style-image-guided, and (3) first-frame-guided video stylization, accompanied by a well-designed data curation pipeline to acquire high-quality paired video data. DreamStyle is built on a vanilla Image-to-Video (I2V) model and trained using a Low-Rank Adaptation (LoRA) with token-specific up matrices that reduces the confusion among different condition tokens. Both qualitative and quantitative evaluations demonstrate that DreamStyle is competent in all three video stylization tasks, and outperforms the competitors in style consistency and video quality. |
| title | DreamStyle: A Unified Framework for Video Stylization |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2601.02785 |