SplitFlux：从单张图像学习解耦内容与风格
SplitFlux: Learning to Decouple Content and Style from a Single Image
对 Flux 单流块做 LoRA 微调并按秩约束与显著性门控拆分内容/风格分支，为视觉研究者提供了一个可直接复用的「从单图分离内容与风格再重嵌入新语境」的定制化生成流程，也可迁移到其他 DiT 架构的可控编辑。
原文依据 we propose SplitFlux, which disentangles content and style by fine-tuning the single stream blocks via LoRA
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论解耦图像内容与风格对于定制化图像生成至关重要。现有的基于 SDXL 的方法难以取得高质量结果，而近期提出的 Flux 模型因其特性尚未被充分探索，无法实现有效的内容-风格分离。为应对这些挑战，我们对 Flux 进行了系统性分析，并得到两点关键观察：(1) 单流块对图像生成至关重要；(2) 早期的单流块主要控制内容，而较晚的块主导风格。基于这些洞见，我们提出 SplitFlux，通过 LoRA 微调单流块来解耦内容与风格，使解耦后的内容能够被重新嵌入新的语境。它包含两个关键组件：(1) 秩约束适配。为保持内容的身份与结构，我们在特定块内压缩秩并放大更新的幅度，防止内容泄漏到风格块中。(2) 视觉门控 LoRA。我们依据图像显著性将内容 LoRA 拆分为两个不同秩的分支。高秩分支保留主体信息，低秩分支编码残差细节，从而缓解内容过拟合并实现无缝重嵌入。大量实验表明，SplitFlux 持续优于最优方法，在多种场景下实现了更优的内容保持与风格化质量。代码可在 https://github.com/yangyt46/SplitFlux 获取。
英文原文摘要 1534 字符
Disentangling image content and style is essential for customized image generation. Existing SDXL-based methods struggle to achieve high-quality results, while the recently proposed Flux model fails to achieve effective content-style separation due to its underexplored characteristics. To address these challenges, we conduct a systematic analysis of Flux and make two key observations: (1) Single Stream Blocks are essential for image generation; and (2) Early single stream blocks mainly control content, whereas later blocks govern style. Based on these insights, we propose SplitFlux, which disentangles content and style by fine-tuning the single stream blocks via LoRA, enabling the disentangled content to be re-embedded into new contexts. It includes two key components: (1) Rank-Constrained Adaptation. To preserve content identity and structure, we compress the rank and amplify the magnitude of updates within specific blocks, preventing content leakage into style blocks. (2) Visual-Gated LoRA. We split the content LoRA into two branches with different ranks, guided by image saliency. The high-rank branch preserves primary subject information, while the low-rank branch encodes residual details, mitigating content overfitting and enabling seamless re-embedding. Extensive experiments demonstrate that SplitFlux consistently outperforms state-of-the-art methods, achieving superior content preservation and stylization quality across diverse scenarios. The code can be accessed at https://github.com/yangyt46/SplitFlux