--- title: ShenDuMao-MOSS-Edit-LoRA-v2 canonical_url: "https://www.modelscope.cn/models/shendumao/ShenDuMao-MOSS-Edit-LoRA-v2" md_url: "https://www.modelscope.cn/models/shendumao/ShenDuMao-MOSS-Edit-LoRA-v2.md" repository: shendumao/ShenDuMao-MOSS-Edit-LoRA-v2 chinese_name: "神都猫 MOSS 单图编辑 LoRA v2 · 2000步" last_updated: 2026-10-20 license: apache-2.0 base_model: - OpenMOSS-Team/MOSS-Image-Base base_model_relation: adapter parameters: 36.3M tensor_type: - BF16 library_name: - safetensors language: - zh - en downloads: 64 stars: 3 tags: - moss-image - lora - image-editing - comfyui --- # ShenDuMao-MOSS-Edit-LoRA-v2 > ShenDuMao-MOSS-Edit-LoRA-v2 - shendumao 在 ModelScope 开源的模型。神都猫使用 AI-Toolkit 训练的 MOSS 单参考图编辑 LoRA，v2，2000步。配套 ComfyUI 插件、SFT 工作流与中英文教程：https://github.com/signerzwb/ComfyUI-ShenDuMao-MOSS-Edit | Trained by ShenDuMao with AI-Toolkit, 2,000 steps. MOSS… shendumao/ShenDuMao-MOSS-Edit-LoRA-v2 是 ModelScope 魔搭社区上的 36.3M 参数机器学习模型，采用 apache-2.0 许可，基于 OpenMOSS-Team/MOSS-Image-Base 构建。 - **Repository**: shendumao/ShenDuMao-MOSS-Edit-LoRA-v2 - **License**: apache-2.0 - **Parameters**: 36.3M - **Base model**: OpenMOSS-Team/MOSS-Image-Base - **Tags**: moss-image, lora, image-editing, comfyui - **Downloads**: 64 - **Stars**: 3 - **Last updated**: 2026-10-20 Source: https://www.modelscope.cn/models/shendumao/ShenDuMao-MOSS-Edit-LoRA-v2 --- # 神都猫 MOSS 单图编辑 LoRA v2 / ShenDuMao MOSS Edit LoRA v2 **神都猫使用 AI-Toolkit 训练，2000 步，v2。** 这是单参考图编辑 LoRA，本页示例使用 **MOSS Image SFT** 推理。 **Trained by ShenDuMao (神都猫) with AI-Toolkit, 2,000 steps, v2.** A single-reference edit LoRA; examples here were generated with **MOSS Image SFT**. ## 插件与下载 / Plugin and downloads - **ComfyUI 插件、安装教程、中英文技术介绍、工作流 / Plugin, installation, bilingual technical notes and workflows:** [ComfyUI-ShenDuMao-MOSS-Edit](https://github.com/signerzwb/ComfyUI-ShenDuMao-MOSS-Edit) - **LoRA 权重 / LoRA weights:** [moss_wendingedit_lr5e-5_1024_v2_000002000.safetensors](https://www.modelscope.cn/models/shendumao/ShenDuMao-MOSS-Edit-LoRA-v2/resolve/master/moss_wendingedit_lr5e-5_1024_v2_000002000.safetensors) - **模型首页 / Model page:** [ModelScope](https://www.modelscope.cn/models/shendumao/ShenDuMao-MOSS-Edit-LoRA-v2) ## 安装和使用 / Installation and usage 1. 更新到最新版 ComfyUI；在 `ComfyUI/custom_nodes/` 执行下方命令并重启。 Update ComfyUI, clone the plugin into `custom_nodes`, and restart. ```bash git clone https://github.com/signerzwb/ComfyUI-ShenDuMao-MOSS-Edit.git ``` 2. 下载本页 LoRA，放到 `ComfyUI/models/loras/`。MOSS SFT 放在 `models/diffusion_models/`，完整 Qwen3.5 视觉/文本编码器放在 `models/text_encoders/`，FLUX.2 VAE 放在 `models/vae/`。 Put this LoRA in `ComfyUI/models/loras/`. Place MOSS SFT in `models/diffusion_models/`, the full Qwen3.5 vision/text encoder in `models/text_encoders/`, and the FLUX.2 VAE in `models/vae/`. 3. 导入 [SFT + v2 工作流](https://github.com/signerzwb/ComfyUI-ShenDuMao-MOSS-Edit/blob/codex/release/workflows/shendumao_moss_sft_v2_edit.json)，选择模型和 LoRA，上传参考图并输入完整编辑指令。编码器必须包含视觉权重，不能是 text-only 版本。 Import the workflow, select the model and LoRA, upload one reference, and enter the full instruction. A full vision encoder is required. 4. 起始参数 / Starting settings: **50 steps, CFG 3, Euler, shift 5, LoRA 1.0, seed 42**, empty negative prompt. 保持原图比例，宽高为 16 的倍数。Use the reference aspect ratio and dimensions divisible by 16. 5. 使用工作流自带的 sigma 节点；Prepare 内已处理负面分支第 22 层跳过，无需再加 SkipLayerGuidance。Keep the supplied sigma node; negative-branch layer-22 skipping is already handled internally. ## 提示词 / Prompt format ```text editpicforpartV1, [describe the specific edit], Only modify the requested content, and try to keep the rest of the image unchanged as much as possible ``` 写清楚数量、位置、原来的外观、目标外观和保留项。Specify counts, locations, existing appearance, target appearance and what should remain. ## 图片案例 / Examples 原图均由 MOSS SFT 生成；编辑结果直接来自 SFT + 本 LoRA。未通过修图、遮罩合成或其他模型修复，PNG 保留实际工作流。References were generated with MOSS SFT. Edited images are direct SFT + v2 outputs without retouching, masks, compositing or another model's repair. PNGs contain the actual workflows. ### 添加金框眼镜 / Add gold-rimmed glasses ![Glasses comparison](https://raw.githubusercontent.com/signerzwb/ComfyUI-ShenDuMao-MOSS-Edit/refs/heads/codex/release/examples/01_portrait_glasses/comparison.jpg) 眼镜添加生效，人物与背景整体保留，脸部和纹理有轻微重绘。The glasses were added with broadly preserved identity and setting; facial and texture details were slightly redrawn. ### 皮包包身改色 / Handbag body recolor ![Handbag comparison](https://raw.githubusercontent.com/signerzwb/ComfyUI-ShenDuMao-MOSS-Edit/refs/heads/codex/release/examples/08_handbag_explicit_whole/comparison.jpg) 包身正面、翻盖、侧面由焦糖棕变雾蓝；金属扣与走线大体保留。提手仍棕色，并未完成所有皮革部位改色。The front, flap and sides changed from caramel brown to powder blue. The clasp and stitching were broadly preserved; the handle remained brown, so not every requested leather part changed. ## 更多类型与指令复测 / More image types and explicit instructions 写真、家居、美食和水彩插画的实际结果如下，同时包含有效修改和未完成项。对照图只做缩放、拼排与加说明，没有修复生成内容。The following selected comparisons include effective edits and remaining limitations; only layout and labels were added. ### 皮包：明确部件后，改色范围更完整 / Handbag: explicit part descriptions improve coverage ![皮包：明确部件后，改色范围更完整](https://raw.githubusercontent.com/signerzwb/ComfyUI-ShenDuMao-MOSS-Edit/refs/heads/codex/release/examples/gallery/02_handbag_prompt_comparison.jpg) 从只改翻盖和侧面，到正面包身也变蓝；金属扣与走线大体保留，提手仍为棕色。 Explicit part descriptions extended the recolor to the front body. The handle still remained brown. ### 写真：白色针织毛衣改蓝 / Fashion photo: recoloring a knitted sweater ![写真：白色针织毛衣改蓝](https://raw.githubusercontent.com/signerzwb/ComfyUI-ShenDuMao-MOSS-Edit/refs/heads/codex/release/examples/gallery/03_knit_sweater.jpg) 毛衣改蓝生效，街景与姿势大体保持；脸部、衣领和针织纹理仍有重绘。 The sweater became blue; pose and street layout broadly stayed, while face, collar and knit details changed. ### 家居：沙发改色与抱枕保留的边界 / Interior: sofa recolor and cushion preservation limits ![家居：沙发改色与抱枕保留的边界](https://raw.githubusercontent.com/signerzwb/ComfyUI-ShenDuMao-MOSS-Edit/refs/heads/codex/release/examples/gallery/04_sofa_prompt_comparison.jpg) 沙发变绿；详细指令改善了左侧黄抱枕，但右侧保留不足，并出现额外绿色物体。 Sofa recolor worked. Yellow-cushion preservation remained incomplete, with an unwanted green object. ### 美食：明确六颗草莓后，仍未完整替换 / Food: six positions targeted, incomplete fruit replacement ![美食：明确六颗草莓后，仍未完整替换](https://raw.githubusercontent.com/signerzwb/ComfyUI-ShenDuMao-MOSS-Edit/refs/heads/codex/release/examples/gallery/05_cake_six_strawberries.jpg) 六处位置均得到处理，但仍有红色草莓果肉和绿叶，呈混合形态，不算完整替换成功。 All six positions were targeted, but strawberry flesh and leaves remained: object replacement was incomplete. ### 水彩插画：斗篷部分改色，背景整体保留 / Watercolor: partial cape recolor and broad scene preservation ![水彩插画：斗篷部分改色，背景整体保留](https://raw.githubusercontent.com/signerzwb/ComfyUI-ShenDuMao-MOSS-Edit/refs/heads/codex/release/examples/gallery/06_watercolor_cape.jpg) 前方衣片变红，外侧仍有浅蓝；详细指令使尾巴保留橘色，但轮廓和姿态仍改变。 Front cape panels became red; blue fabric remained outside. Orange tail color improved, but its shape changed. ### 工作流分区 / Workflow layout ![Organized workflow](https://raw.githubusercontent.com/signerzwb/ComfyUI-ShenDuMao-MOSS-Edit/refs/heads/codex/release/examples/gallery/07_workflow.jpg) [下载更多对照图库 / Download the comparison gallery](https://www.modelscope.cn/models/shendumao/ShenDuMao-MOSS-Edit-LoRA-v2/resolve/master/MOSS-v2-more-before-after-comparisons.zip) ## 技术依据 / Technical basis 配套插件参考 [Ostris AI-Toolkit](https://github.com/ostris/ai-toolkit) 的 MOSS 编辑推理约定，实现两路图像条件、隔离参考注意力、采样期 K/V 缓存和对齐的 sigma 调度。Plugin details and upstream attribution are documented in the GitHub repository. 当前为实验性编辑权重；局部细节可能重绘，数量、物体替换、背景更换等指令不保证完整完成。This is experimental: non-target details may be redrawn, and counting, object replacement or background edits may not be completed. ## 署名与许可 / Credits and license **© 2026 神都猫 / ShenDuMao. Trained with AI-Toolkit, 2,000 steps, v2.** LoRA: Apache License 2.0. 配套插件代码为 MIT，并保留 Ostris 的原始署名。Plugin code is MIT with upstream attribution retained. MOSS 底模与其他组件分别遵循各自许可。Base models and other components retain their respective licenses.