ModelsLab/RedCraft | 红潮 CADS | UPdated-May11 | Commercial & Advertising Design System - RED-Edit (ICEdit)1.1🔥多任务

5.11 更新RED-Edit v1.1(基于ICEdit normal LoRA) 多指令优化

5.11 右侧下载列表中的 Trainning data 工作流更新

---

ICEdit 多模态控制器(IC引导下的图像编辑)

In-Context Edit: Enabling Instructional Image Editing with In-Context Generation in Large Scale Diffusion Transformer

基于 Flux.tools-Fill 最低仅需 6G 显存

近期出现的多模态控制器对C端用户越来越不友好

动不动就要20-30+显存，望而却步

ICEdit 让我们重新相信光

---

RED-Edit 是在 RED-Fill (NSFW) 基础上合并了 ICEdit 训练权重

最低仅需 8步 推理

工作流及模型文件在下载列表中

工作流打包为 "Trainning data" 压缩文件

---

感谢： @river-zhang 及团队成员 浙江大学  & Harvard University

---

Compared with commercial models such as Gemini and GPT-4O, our methods are comparable to and even superior to these commercial models in terms of character ID preservation and instruction following. We are more open-source than them, with lower costs, faster speed (it takes about 9 seconds to process one image), and powerful performance.

---

进阶玩法1 Advanced 1：叠加 Flux.fill LoRas

可以获得更稳定的修改结果：

[Flux Fill] Virtual Try Off 虚拟脱衣

@xiaozhijason / [Flux Fill] Virtual Try Off - alpha

Object Removal Flux Fill v2 物体移除

@xiaozhijason / Object Removal Flux Fill v2 - v2.0

---

进阶玩法2 Advanced 2： 叠加 Flux IC-LoRA

虚拟换装 In-Context 模型地址：

xiaozaa/catvton-flux-lora-alpha at main

感谢 T8x 提到的这个方案，十字鱼 提供 整合包 下载

---

@article{zhang2025ICEdit,
  title={In-Context Edit: Enabling Instructional Image Editing with In-Context Generation in Large Scale Diffusion Transformer},
  author={Zhang, Zechuan and Xie, Ji and Lu, Yu and Yang, Zongxin and Yang, Yi},
  journal={arXiv},
  year={2025},
  url={https://arxiv.org/abs/2504.20690},
}