Storybooth: Training-free Multi-Subject Consistency for Improved Visual Storytelling
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908306762629120 |
|---|---|
| author | Singh, Jaskirat Chen, Junshen Kevin Kohler, Jonas Cohen, Michael |
| author_facet | Singh, Jaskirat Chen, Junshen Kevin Kohler, Jonas Cohen, Michael |
| contents | Training-free consistent text-to-image generation depicting the same subjects across different images is a topic of widespread recent interest. Existing works in this direction predominantly rely on cross-frame self-attention; which improves subject-consistency by allowing tokens in each frame to pay attention to tokens in other frames during self-attention computation. While useful for single subjects, we find that it struggles when scaling to multiple characters. In this work, we first analyze the reason for these limitations. Our exploration reveals that the primary-issue stems from self-attention-leakage, which is exacerbated when trying to ensure consistency across multiple-characters. This happens when tokens from one subject pay attention to other characters, causing them to appear like each other (e.g., a dog appearing like a duck). Motivated by these findings, we propose StoryBooth: a training-free approach for improving multi-character consistency. In particular, we first leverage multi-modal chain-of-thought reasoning and region-based generation to apriori localize the different subjects across the desired story outputs. The final outputs are then generated using a modified diffusion model which consists of two novel layers: 1) a bounded cross-frame self-attention layer for reducing inter-character attention leakage, and 2) token-merging layer for improving consistency of fine-grain subject details. Through both qualitative and quantitative results we find that the proposed approach surpasses prior state-of-the-art, exhibiting improved consistency across both multiple-characters and fine-grain subject details. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_05800 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Storybooth: Training-free Multi-Subject Consistency for Improved Visual Storytelling Singh, Jaskirat Chen, Junshen Kevin Kohler, Jonas Cohen, Michael Computer Vision and Pattern Recognition Machine Learning Multimedia Training-free consistent text-to-image generation depicting the same subjects across different images is a topic of widespread recent interest. Existing works in this direction predominantly rely on cross-frame self-attention; which improves subject-consistency by allowing tokens in each frame to pay attention to tokens in other frames during self-attention computation. While useful for single subjects, we find that it struggles when scaling to multiple characters. In this work, we first analyze the reason for these limitations. Our exploration reveals that the primary-issue stems from self-attention-leakage, which is exacerbated when trying to ensure consistency across multiple-characters. This happens when tokens from one subject pay attention to other characters, causing them to appear like each other (e.g., a dog appearing like a duck). Motivated by these findings, we propose StoryBooth: a training-free approach for improving multi-character consistency. In particular, we first leverage multi-modal chain-of-thought reasoning and region-based generation to apriori localize the different subjects across the desired story outputs. The final outputs are then generated using a modified diffusion model which consists of two novel layers: 1) a bounded cross-frame self-attention layer for reducing inter-character attention leakage, and 2) token-merging layer for improving consistency of fine-grain subject details. Through both qualitative and quantitative results we find that the proposed approach surpasses prior state-of-the-art, exhibiting improved consistency across both multiple-characters and fine-grain subject details. |
| title | Storybooth: Training-free Multi-Subject Consistency for Improved Visual Storytelling |
| topic | Computer Vision and Pattern Recognition Machine Learning Multimedia |
| url | https://arxiv.org/abs/2504.05800 |