通过分离身份与场景嵌入，提升多图故事生成的一致性
ReDiStory: Region-Disentangled Diffusion for Consistent Visual Story Generation
- 将文本嵌入分解为身份与帧特异性两部分，减少跨帧干扰
- 在相同模型下，身份一致性指标显著优于基线方法
- 无需训练或额外监督，适合快速部署于扩散模型
生成连贯的视觉故事需在多张图像间保持主体身份一致并保留每帧语义。现有无训练方法将身份与帧提示拼接为统一表示，常引发跨帧语义干扰，削弱身份保持能力。本文提出ReDiStory，一种推理时进行提示嵌入重组的无训练框架。该方法显式将文本嵌入分解为与身份相关和帧特定的成分，并通过抑制跨帧共享方向来解耦帧嵌入，从而降低跨帧干扰，且不修改扩散参数或需要额外监督。在相同扩散主干和推理设置下，ReDiStory在保持提示忠实度的同时提升了身份一致性。在ConsiStory+基准测试中，多个身份一致性指标均优于1Prompt1Story。代码已开源：https://github.com/YuZhenyuLindy/ReDiStory
原文摘要 · Abstract (English)
Generating coherent visual stories requires maintaining subject identity across multiple images while preserving frame-specific semantics. Recent training-free methods concatenate identity and frame prompts into a unified representation, but this often introduces inter-frame semantic interference that weakens identity preservation in complex stories. We propose ReDiStory, a training-free framework that improves multi-frame story generation via inference-time prompt embedding reorganization. ReDiStory explicitly decomposes text embeddings into identity-related and frame-specific components, then decorrelates frame embeddings by suppressing shared directions across frames. This reduces cross-frame interference without modifying diffusion parameters or requiring additional supervision. Under identical diffusion backbones and inference settings, ReDiStory improves identity consistency while maintaining prompt fidelity. Experiments on the ConsiStory+ benchmark show consistent gains over 1Prompt1Story on multiple identity consistency metrics. Code is available at: https://github.com/YuZhenyuLindy/ReDiStory
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。