让扩散模型在换风格时保持细节一致，还能避免风格变差。
OmniConsistency: Learning Style-Agnostic Consistency from Paired Stylization Data
- 用成对图像训练一致性，让模型学会跨场景保持风格统一。
- 分两阶段学习，先学风格再保一致，防止风格退化。
- 插件式设计，兼容任意风格LoRA，适合想稳定出图的用户。
扩散模型显著推动了图像风格化发展，但仍面临两大挑战：（1）复杂场景中保持身份、构图和细节点的一致性；（2）在使用风格LoRA的图像到图像流水线中防止风格退化。GPT-4o展现出极高的风格一致性，凸显开源方法与商用模型之间的差距。为此，我们提出 extbf{OmniConsistency}，一种通用一致性插件，基于大规模扩散Transformer（DiTs）。其贡献包括：（1）基于对齐图像对的上下文一致性学习框架，实现鲁棒泛化；（2）两阶段渐进式学习策略，解耦风格学习与一致性保留，缓解风格退化；（3）完全即插即用设计，兼容任意风格LoRA，适用于Flux框架。大量实验表明，OmniConsistency显著提升视觉连贯性和美学质量，性能接近商业顶级模型GPT-4o。
原文摘要 · Abstract (English)
Diffusion models have advanced image stylization significantly, yet two core challenges persist: (1) maintaining consistent stylization in complex scenes, particularly identity, composition, and fine details, and (2) preventing style degradation in image-to-image pipelines with style LoRAs. GPT-4o's exceptional stylization consistency highlights the performance gap between open-source methods and proprietary models. To bridge this gap, we propose \textbf{OmniConsistency}, a universal consistency plugin leveraging large-scale Diffusion Transformers (DiTs). OmniConsistency contributes: (1) an in-context consistency learning framework trained on aligned image pairs for robust generalization; (2) a two-stage progressive learning strategy decoupling style learning from consistency preservation to mitigate style degradation; and (3) a fully plug-and-play design compatible with arbitrary style LoRAs under the Flux framework. Extensive experiments show that OmniConsistency significantly enhances visual coherence and aesthetic quality, achieving performance comparable to commercial state-of-the-art model GPT-4o.
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。