license: apache-2.0# Anime Image Purifier (Kontext LoRA)This is a LoRA model designed to clean and enhance anime-style images by removing watermarks, text, and other visual artifacts. ## 📖 Model DescriptionThis model was trained using the [kohya-ss/musubi-tuner](<a target="_blank" rel="ugc" href="https://github.com/kohya-ss/musubi-tuner)**">https://github.com/kohya-ss/musubi-tuner)</a> script on a dataset of 21,920 anime images. Its primary purpose is to improve image quality, remove unwanted text and watermarks, and clean up the overall picture. The model can effectively process images with distracting text or logos to generate a clean, high-quality version.### 📖 基本说明本模型使用 [kohya-ss/musubi-tuner](<a target="_blank" rel="ugc" href="https://github.com/kohya-ss/musubi-tuner)**">https://github.com/kohya-ss/musubi-tuner)</a> 训练脚本，通过 21,920 对动漫图片进行训练。其主要目的是修复图像质量、移除图片中不需要的文字和水印，并清理整体画面。该模型能够有效地处理带有干扰性文本或标志的图像，生成一个干净、清晰的版本。A new V2 version of this model is now available and highly recommended. Compared to V1, the V2 model offers significantly improved performance, capable of addressing a wider and more complex range of image degradations. This includes enhancements in handling distortions related to color, contrast, and resolution, as well as the removal of random, obstructive stickers. For superior results, please use the V2 version.模型的V2版本现已发布，我们强烈推荐您直接使用。 与V1版本相比，V2版本在性能上有了显著提升，能够应对更为复杂的图像破坏，包括色彩、对比度、分辨率的失真，以及随机贴纸的干扰。为了获得最佳的修复效果，请优先选择V2版本。## 🚀 How to UseIn a compatible workflow (such as the official Kontext workflow in ComfyUI), add a "Load LoRA" node and select this model. Then, add the following trigger phrase to your Positive Prompt to activate the cleaning and restoration effect:```clean image, remove watermark and text````### 🚀 基本使用在兼容的工作流（如ComfyUI官方的Kontext工作流）中，增加一个加载Lora节点并选择本Lora模型。随后，在您的正向提示词 (Positive Prompt) 中加入触发词 <code>"clean image, remove watermark and text"</code> 即可引导模型执行清洁和修复任务。```clean image, remove watermark and text````---## 🖼️ Showcase / 案例The left side shows the original image, and the right side shows the cleaned version after applying the model.左侧为原始待处理图像，右侧为经过本模型修复后的高清无水印文字版本。 ---## ⚠️ Important NoticeWhen using this model, please respect the copyright of the original images. This tool is intended for academic research, personal projects, or for processing images you have the legal right to use. Do not use it for illegal purposes or to infringe upon the rights of others.### ⚠️ 注意事项请注意，使用本模型时，请尊重原始图像的版权。此工具旨在用于学术研究、个人项目或处理您拥有合法使用权的图像。**请勿用于非法目的或侵犯他人权益。**

Files with SHA256: 224ba523f00ae80a0da853b50c1360d6c29f01e88010b0073ee42ac9d3152beb