[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-chaiNNer-org--chaiNNer":3,"tool-chaiNNer-org--chaiNNer":64},[4,17,27,35,48,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",138956,2,"2026-04-05T11:33:21",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,43,44,45,15,46,26,13,47],"数据工具","视频","插件","其他","音频",{"id":49,"name":50,"github_repo":51,"description_zh":52,"stars":53,"difficulty_score":10,"last_commit_at":54,"category_tags":55,"status":16},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,"2026-04-04T04:44:48",[15,14,13,26,46],{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"status":16},519,"PaddleOCR","PaddlePaddle\u002FPaddleOCR","PaddleOCR 是一款基于百度飞桨框架开发的高性能开源光学字符识别工具包。它的核心能力是将图片、PDF 等文档中的文字提取出来，转换成计算机可读取的结构化数据，让机器真正“看懂”图文内容。\n\n面对海量纸质或电子文档，PaddleOCR 解决了人工录入效率低、数字化成本高的问题。尤其在人工智能领域，它扮演着连接图像与大型语言模型（LLM）的桥梁角色，能将视觉信息直接转化为文本输入，助力智能问答、文档分析等应用场景落地。\n\nPaddleOCR 适合开发者、算法研究人员以及有文档自动化需求的普通用户。其技术优势十分明显：不仅支持全球 100 多种语言的识别，还能在 Windows、Linux、macOS 等多个系统上运行，并灵活适配 CPU、GPU、NPU 等各类硬件。作为一个轻量级且社区活跃的开源项目，PaddleOCR 既能满足快速集成的需求，也能支撑前沿的视觉语言研究，是处理文字识别任务的理想选择。",74913,"2026-04-05T10:44:17",[26,14,13,46],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":67,"owner_avatar_url":76,"owner_bio":77,"owner_company":78,"owner_location":78,"owner_email":78,"owner_twitter":78,"owner_website":79,"owner_url":80,"languages":81,"stars":105,"forks":106,"last_commit_at":107,"license":108,"difficulty_score":23,"env_os":109,"env_gpu":110,"env_ram":111,"env_deps":112,"category_tags":121,"github_topics":78,"view_count":23,"oss_zip_url":78,"oss_zip_packed_at":78,"status":16,"created_at":122,"updated_at":123,"faqs":124,"releases":155},3923,"chaiNNer-org\u002FchaiNNer","chaiNNer","A node-based image processing GUI aimed at making chaining image processing tasks easy and customizable. Born as an AI upscaling application, chaiNNer has grown into an extremely flexible and powerful programmatic image processing application.","chaiNNer 是一款基于节点流程的图像处理图形界面工具，旨在让复杂的图像任务变得像搭积木一样简单直观。它最初专为 AI 图像超分辨率（放大）设计，如今已进化为功能极其强大的通用图像处理平台。\n\n对于需要精细控制工作流的用户而言，chaiNNer 解决了传统软件操作僵化、难以定制复杂处理链条的痛点。用户只需通过拖拽节点并连线，即可自由组合出从基础调整到高度复杂的自动化处理流程，无需编写任何代码。\n\n这款工具非常适合设计师、AI 爱好者以及希望尝试深度学习模型但畏惧命令行操作的普通用户。无论是进行图片高清修复、格式转换还是批量处理，都能轻松上手。同时，其开放的架构也吸引了开发者和研究人员参与扩展。\n\n技术亮点方面，chaiNNer 支持跨平台运行（Windows、macOS、Linux），并内置了独立的 Python 环境，用户无需手动配置繁琐的开发依赖。它广泛兼容 PyTorch、NCNN、ONNX 和 TensorRT 等主流神经网络框架，能根据用户的显卡类型（如 NVIDIA 或 AMD）智能推荐最优方案，让高性能 AI 推理触手可及。","# chaiNNer\n\n[![GitHub Latest Release](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fv\u002Frelease\u002FchaiNNer-org\u002FchaiNNer)](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Freleases\u002Flatest)\n[![GitHub Total Downloads](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002FchaiNNer-org\u002FchaiNNer\u002Ftotal)](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Freleases)\n[![License](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002FchaiNNer-org\u002FchaiNNer)](.\u002FLICENSE)\n[![Discord](https:\u002F\u002Fimg.shields.io\u002Fdiscord\u002F930865462852591648?label=Discord&logo=Discord&logoColor=white&color=5865F2)](https:\u002F\u002Fdiscord.gg\u002FpzvAKPKyHM)\n[![ko-fi](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FKo--fi-Support%20chaiNNer%20-hotpink?logo=kofi&logoColor=white)](https:\u002F\u002Fko-fi.com\u002FT6T46KTTW)\n![PRs Welcome](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPRs-welcome-brightgreen)\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Freleases\" target=\"_blank\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FchaiNNer-org_chaiNNer_readme_a7c119be1ad6.png\" width=\"720\" \u002F>\n  \u003C\u002Fa>\n\u003C\u002Fp>\n\nA node-based image processing GUI aimed at making chaining image processing tasks easy and customizable. Born as an AI upscaling application, chaiNNer has grown into an extremely flexible and powerful programmatic image processing application.\n\nChaiNNer gives you a level of customization of your image processing workflow that very few others do. Not only do you have full control over your processing pipeline, you can do incredibly complex tasks just by connecting a few nodes together.\n\nChaiNNer is also cross-platform, meaning you can run it on Windows, MacOS, and Linux.\n\nFor help, suggestions, or just to hang out, you can join the [chaiNNer Discord server](https:\u002F\u002Fdiscord.gg\u002FpzvAKPKyHM)\n\nChaiNNer is under active development. If you're knowledgeable in TypeScript, React, or Python, feel free to contribute to this project and help us continue to improve it.\n\n## Installation\n\nDownload the latest release from the [Github releases page](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Freleases) and run the installer best suited for your system. Simple as that.\n\nYou don't even need to have Python installed, as chaiNNer will download an isolated integrated Python build on startup. From there, you can install all the other dependencies via the Dependency Manager.\n\nIf you do wish to use your system Python installation still, you can turn the system Python setting on. However, it is much more recommended to use integrated Python. If you do wish to use your system Python, Python 3.10 or later is required (3.11+ recommended).\n\nIf you'd like to test the latest changes and tweaks, try out our [nightly builds](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer-nightly)\n\n## How To Use\n\n### Basic Usage\n\nWhile it might seem intimidating at first due to all the possible options, chaiNNer is pretty simple to use. For example, this is all you need to do in order to perform an upscale:\n\n\u003Cp align=\"center\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FchaiNNer-org_chaiNNer_readme_81004277a445.png\" width=\"480\" \u002F>\n\u003C\u002Fp>\n\nBefore you get to this point though, you'll need to install one of the neural network frameworks from the dependency manager. You can access this via the button in the upper-right-hand corner. ChaiNNer offers support for PyTorch (with select model architectures), NCNN, ONNX, and TensorRT. For Nvidia users, PyTorch or TensorRT will be the preferred way to upscale. For AMD users, NCNN will be the preferred way to upscale (or PyTorch with ROCm on Linux).\n\nAll the other Python dependencies are automatically installed, and chaiNNer even carries its own integrated Python support so that you do not have to modify your existing Python configuration.\n\nThen, all you have to do is drag and drop (or double click) node names in the selection panel to bring them into the editor. Then, drag from one node handle to another to connect the nodes. Each handle is color-coded to its specific type, and while connecting will show you only the compatible connections. This makes it very easy to know what to connect where.\n\nOnce you have a working chain set up in the editor, you can press the green \"run\" button in the top bar to run the chain you have made. You will see the connections between nodes become animated, and start to un-animate as they finish processing. You can stop or pause processing with the red \"stop\" and yellow \"pause\" buttons respectively.\n\n\u003Cp align=\"center\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FchaiNNer-org_chaiNNer_readme_dd3daa85cb98.png\" width=\"540\" \u002F>\n\u003C\u002Fp>\n\nDon't forget, there are plenty of non-upscaling tasks you can do with chaiNNer as well!\n\n### Tips & Tricks\n\nTo select multiple nodes, hold down shift and drag around all the nodes you want to be selected. You can also select an individual node by just clicking on it. When nodes are selected, you can press backspace or delete to delete them from the editor.\n\nTo perform batch processing on a folder of images, use the \"Load Images\" node. To process videos, use the \"Load Video\" node. It's important to note however that you cannot use both \"Load Images\" and \"Load Video\" nodes (or any two nodes that perform batch iteration) together in a chain. You can however combine the output (collector) nodes in the chain, for example using \"Save Image\" with \"Load Video\", and \"Save Video\" with \"Load Images\".\n\nYou can right-click in the editor viewport to show an inline nodes list to select from. You also can get this menu by dragging a connection out to the editor rather than making an actual connection, and it will show compatible nodes to automatically create a connection with.\n\n### Helpful Resources\n\n-   [Kim's chaiNNer Templates](https:\u002F\u002Fgithub.com\u002Fkimberly990\u002Fkim-chaiNNer-Templates\u002F)\n    -   A collection of useful chain templates that can quickly get you started if you are still new to using chaiNNer.\n-   [OpenModelDB Model Database](https:\u002F\u002Fopenmodeldb.info\u002F)\n    -   A nice collection of Super-Resolution models that have been trained by the community.\n-   [Interactive Visual Comparison of Upscaling Models](https:\u002F\u002Fphhofm.github.io\u002Fupscale\u002Fmultimodels.html)\n    -   An online comparison of different models. The author also provides a list of [favorites](https:\u002F\u002Fphhofm.github.io\u002Fupscale\u002Ffavorites.html).\n\n## Compatibility Notes\n\n-   MacOS versions 10.x and below are not supported.\n\n-   Windows versions 8.1 and below are also not supported.\n\n-   Apple Silicon Macs are supported with PyTorch MPS acceleration. ONNX only supports the CPU Execution Provider, and NCNN may not work properly on some configurations.\n\n-   Some NCNN users with non-Nvidia GPUs might get all-black outputs. I am not sure what to do to fix this as it appears to be due to the graphics driver crashing as a result of going out of memory. If this happens to you, try manually setting a tiling amount.\n\n-   To use the Clipboard nodes, Linux users need to have xclip or, for wayland users, wl-copy installed.\n\n## GPU Support\n\n**Nvidia GPUs:** Full support via PyTorch (CUDA), ONNX, and TensorRT. TensorRT offers the best performance for supported models.\n\n**AMD GPUs:**\n- On Linux, AMD GPUs can use PyTorch via ROCm\n- NCNN is available on all platforms for AMD GPUs\n\n**Apple Silicon (M1\u002FM2\u002FM3):** PyTorch MPS acceleration is supported.\n\n**Intel GPUs:** NCNN inference is supported for Intel GPUs.\n\n**CPU:** All frameworks support CPU-only mode as a fallback.\n\nFor NCNN, make sure to select which GPU you want to use in the settings. It might be defaulting to your integrated graphics!\n\n## Model Architecture Support\n\nChaiNNer currently supports a limited amount of neural network architectures. More architectures will be supported in the future.\n\n### PyTorch\n\nAs of v0.21.0, chaiNNer uses our new package called [Spandrel](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002Fspandrel) to support Pytorch model architectures. For a list of what's supported, [check out the list there](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002Fspandrel#model-architecture-support).\n\n### NCNN\n\n#### Single Image Super Resolution\n\n-   Technically, almost any SR model should work assuming they follow a typical CNN-based SR structure. However, I have only tested with ESRGAN (and its variants) and with Waifu2x.\n\n### ONNX\n\n#### Single Image Super Resolution\n\n-   Similarly to NCNN, technically almost any SR model should work assuming they follow a typical CNN-based SR structure. However, I have only tested with ESRGAN.\n\n#### Background Removal\n\n-   [u2net](https:\u002F\u002Fgithub.com\u002Fdanielgatis\u002Frembg) | [u2net](https:\u002F\u002Fgithub.com\u002Fdanielgatis\u002Frembg\u002Freleases\u002Fdownload\u002Fv0.0.0\u002Fu2net.onnx), [u2netp](https:\u002F\u002Fgithub.com\u002Fdanielgatis\u002Frembg\u002Freleases\u002Fdownload\u002Fv0.0.0\u002Fu2netp.onnx), [u2net_cloth_seg](https:\u002F\u002Fgithub.com\u002Fdanielgatis\u002Frembg\u002Freleases\u002Fdownload\u002Fv0.0.0\u002Fu2net_cloth_seg.onnx), [u2net_human_seg](https:\u002F\u002Fgithub.com\u002Fdanielgatis\u002Frembg\u002Freleases\u002Fdownload\u002Fv0.0.0\u002Fu2net_human_seg.onnx), [silueta](https:\u002F\u002Fgithub.com\u002Fdanielgatis\u002Frembg\u002Freleases\u002Fdownload\u002Fv0.0.0\u002Fsilueta.onnx)\n-   [isnet](https:\u002F\u002Fgithub.com\u002Fxuebinqin\u002FDIS) | [isnet](https:\u002F\u002Fgithub.com\u002Fdanielgatis\u002Frembg\u002Freleases\u002Fdownload\u002Fv0.0.0\u002Fisnet-general-use.onnx)\n\n### TensorRT\n\nTensorRT provides optimized inference for Nvidia GPUs. Models must be converted to TensorRT engine format for use. This offers the best performance on supported hardware.\n\n## Troubleshooting\n\nFor troubleshooting information, view the [troubleshooting document](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fwiki\u002F06--Troubleshooting).\n\n## Building chaiNNer Yourself\n\nI provide pre-built versions of chaiNNer here on GitHub. However, if you would like to build chaiNNer yourself, simply run `npm install` (make sure that you have at least npm v7 installed) to install all the nodejs dependencies, and `npm run make` to build the application.\n\n## FAQ\n\nFor FAQ information, view the [FAQ document](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fwiki\u002F07--FAQ).\n\n## Documentation\n\nFor in-depth documentation covering various aspects of ChaiNNer, including CLI usage, data representation, and a contributor's guide, kindly refer to our [ChaiNNer Wiki](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fwiki).\n\n\n","# chaiNNer\n\n[![GitHub 最新发布](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fv\u002Frelease\u002FchaiNNer-org\u002FchaiNNer)](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Freleases\u002Flatest)\n[![GitHub 总下载量](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002FchaiNNer-org\u002FchaiNNer\u002Ftotal)](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Freleases)\n[![许可证](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002FchaiNNer-org\u002FchaiNNer)](.\u002FLICENSE)\n[![Discord](https:\u002F\u002Fimg.shields.io\u002Fdiscord\u002F930865462852591648?label=Discord&logo=Discord&logoColor=white&color=5865F2)](https:\u002F\u002Fdiscord.gg\u002FpzvAKPKyHM)\n[![ko-fi](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FKo--fi-支持 chaiNNer -hotpink?logo=kofi&logoColor=white)](https:\u002F\u002Fko-fi.com\u002FT6T46KTTW)\n![欢迎提交 PR](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPRs-welcome-brightgreen)\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Freleases\" target=\"_blank\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FchaiNNer-org_chaiNNer_readme_a7c119be1ad6.png\" width=\"720\" \u002F>\n  \u003C\u002Fa>\n\u003C\u002Fp>\n\nchaiNNer 是一款基于节点的图像处理 GUI，旨在让图像处理任务的串联变得简单且可定制。最初作为一款 AI 超分辨率应用诞生，如今 chaiNNer 已发展成为一款极其灵活且功能强大的程序化图像处理工具。\n\nchaiNNer 让你对图像处理工作流拥有极高的自定义能力，这是很少有其他工具能做到的。你不仅可以完全掌控整个处理流程，只需通过连接几个节点就能完成极其复杂的任务。\n\n此外，chaiNNer 还是跨平台的，这意味着你可以在 Windows、macOS 和 Linux 上运行它。\n\n如需帮助、提出建议或只是想聊聊天，欢迎加入 [chaiNNer Discord 社区](https:\u002F\u002Fdiscord.gg\u002FpzvAKPKyHM)。\n\nchaiNNer 目前仍在积极开发中。如果你熟悉 TypeScript、React 或 Python，欢迎为本项目贡献力量，帮助我们持续改进！\n\n## 安装\n\n从 [GitHub 发布页面](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Freleases) 下载最新版本，并运行最适合你系统的安装程序即可。非常简单。\n\n你甚至无需预先安装 Python，因为 chaiNNer 在启动时会自动下载一个独立的集成式 Python 版本。之后，你可以通过依赖管理器安装所有其他依赖项。\n\n如果你仍然希望使用系统自带的 Python 环境，可以启用“使用系统 Python”选项。不过，强烈建议使用集成式 Python。若选择使用系统 Python，则需要 Python 3.10 或更高版本（推荐 3.11 及以上）。\n\n如果你想体验最新的更改和优化，可以尝试我们的 [夜间构建版本](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer-nightly)。\n\n## 使用方法\n\n### 基本用法\n\n尽管由于选项众多，chaiNNer 初次使用时可能会让人感到有些复杂，但实际上它非常易于上手。例如，要进行图像超分辨率处理，你只需执行以下步骤：\n\n\u003Cp align=\"center\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FchaiNNer-org_chaiNNer_readme_81004277a445.png\" width=\"480\" \u002F>\n\u003C\u002Fp>\n\n不过，在此之前，你需要先通过依赖管理器安装其中一个神经网络框架。可以通过右上角的按钮访问该功能。chaiNNer 支持 PyTorch（部分模型架构）、NCNN、ONNX 和 TensorRT。对于 NVIDIA 用户，推荐使用 PyTorch 或 TensorRT 进行超分辨率；而对于 AMD 用户，则推荐使用 NCNN（或在 Linux 上搭配 ROCm 的 PyTorch）。\n\n所有其他 Python 依赖项都会自动安装，而且 chaiNNer 自带集成式 Python 支持，因此你无需修改现有的 Python 配置。\n\n接下来，只需将选择面板中的节点名称拖放到编辑器中即可。然后，从一个节点的端口拖动到另一个节点的端口以建立连接。每个端口都按类型进行了颜色编码，连接时只会显示兼容的连接方式，这使得你很容易知道应该将哪些节点连接在一起。\n\n当编辑器中搭建好工作链后，点击顶部栏的绿色“运行”按钮即可开始执行。你会看到节点之间的连接会动态显示，处理完成后则会恢复原状。你可以分别使用红色“停止”按钮和黄色“暂停”按钮来停止或暂停处理过程。\n\n\u003Cp align=\"center\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FchaiNNer-org_chaiNNer_readme_dd3daa85cb98.png\" width=\"540\" \u002F>\n\u003C\u002Fp>\n\n别忘了，chaiNNer 还可以用于许多非超分辨率的图像处理任务哦！\n\n### 小贴士与技巧\n\n要选择多个节点，请按住 Shift 键并拖动以框选所有需要选择的节点。你也可以直接单击某个节点将其单独选中。选中节点后，按下 Backspace 或 Delete 键即可将其从编辑器中删除。\n\n要对文件夹中的多张图片进行批量处理，可以使用“加载图片”节点；如果要处理视频，则使用“加载视频”节点。需要注意的是，你不能在同一链条中同时使用“加载图片”和“加载视频”节点（或任何两个执行批量迭代的节点）。不过，你可以在链条中组合输出（收集）节点，例如将“保存图片”与“加载视频”结合使用，或将“保存视频”与“加载图片”结合使用。\n\n在编辑器视图中右键单击，会弹出一个内联节点列表供你选择。你也可以通过将连接线拖到编辑器中而非实际建立连接的方式来调出该菜单，此时会显示可自动创建连接的兼容节点。\n\n### 有用资源\n\n-   [Kim 的 chaiNNer 模板库](https:\u002F\u002Fgithub.com\u002Fkimberly990\u002Fkim-chaiNNer-Templates\u002F)\n    -  一系列实用的工作链模板，适合刚接触 chaiNNer 的用户快速上手。\n-   [OpenModelDB 模型数据库](https:\u002F\u002Fopenmodeldb.info\u002F)\n    -  由社区训练的一系列优秀的超分辨率模型集合。\n-   [超分辨率模型交互式可视化对比](https:\u002F\u002Fphhofm.github.io\u002Fupscale\u002Fmultimodels.html)\n    -  一个在线的模型对比工具。作者还提供了一份 [精选模型列表](https:\u002F\u002Fphhofm.github.io\u002Fupscale\u002Ffavorites.html)。\n\n## 兼容性说明\n\n-   不支持 macOS 10.x 及更早版本。\n\n-   不支持 Windows 8.1 及更早版本。\n\n-   Apple Silicon Mac 通过 PyTorch MPS 加速得到支持。ONNX 仅支持 CPU 执行提供者，而 NCNN 在某些配置下可能无法正常工作。\n\n-   部分使用非 NVIDIA 显卡的 NCNN 用户可能会遇到全黑输出的问题。目前尚不清楚如何解决这个问题，但似乎是由显卡驱动因内存不足而崩溃导致的。如果出现这种情况，可以尝试手动设置分块大小。\n\n-   对于 Linux 用户，要使用剪贴板相关节点，需要安装 xclip，或者对于 Wayland 用户，需要安装 wl-copy。\n\n## GPU 支持\n\n**Nvidia GPU：** 通过 PyTorch（CUDA）、ONNX 和 TensorRT 提供全面支持。对于受支持的模型，TensorRT 能够提供最佳性能。\n\n**AMD GPU：**\n- 在 Linux 系统上，AMD GPU 可以通过 ROCm 使用 PyTorch。\n- NCNN 在所有平台上均支持 AMD GPU。\n\n**Apple Silicon（M1\u002FM2\u002FM3）：** 支持 PyTorch MPS 加速。\n\n**Intel GPU：** Intel GPU 支持 NCNN 推理。\n\n**CPU：** 所有框架都支持仅使用 CPU 的模式作为备用方案。\n\n对于 NCNN，请确保在设置中选择您想要使用的 GPU。它可能会默认使用您的集成显卡！\n\n## 模型架构支持\n\nChaiNNer 目前仅支持有限数量的神经网络架构。未来将支持更多架构。\n\n### PyTorch\n\n自 v0.21.0 起，ChaiNNer 使用我们新推出的名为 [Spandrel](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002Fspandrel) 的包来支持 PyTorch 模型架构。有关支持的列表，请查看 [此处](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002Fspandrel#model-architecture-support)。\n\n### NCNN\n\n#### 单张图像超分辨率\n\n- 从技术上讲，只要遵循典型的基于 CNN 的超分辨率结构，几乎任何 SR 模型都应该可以工作。不过，我目前只测试过 ESRGAN（及其变体）和 Waifu2x。\n\n### ONNX\n\n#### 单张图像超分辨率\n\n- 与 NCNN 类似，理论上只要遵循典型的基于 CNN 的超分辨率结构，几乎任何 SR 模型都可以使用。不过，我目前只测试过 ESRGAN。\n\n#### 背景去除\n\n- [u2net](https:\u002F\u002Fgithub.com\u002Fdanielgatis\u002Frembg) | [u2net](https:\u002F\u002Fgithub.com\u002Fdanielgatis\u002Frembg\u002Freleases\u002Fdownload\u002Fv0.0.0\u002Fu2net.onnx)、[u2netp](https:\u002F\u002Fgithub.com\u002Fdanielgatis\u002Frembg\u002Freleases\u002Fdownload\u002Fv0.0.0\u002Fu2netp.onnx)、[u2net_cloth_seg](https:\u002F\u002Fgithub.com\u002Fdanielgatis\u002Frembg\u002Freleases\u002Fdownload\u002Fv0.0.0\u002Fu2net_cloth_seg.onnx)、[u2net_human_seg](https:\u002F\u002Fgithub.com\u002Fdanielgatis\u002Frembg\u002Freleases\u002Fdownload\u002Fv0.0.0\u002Fu2net_human_seg.onnx)、[silueta](https:\u002F\u002Fgithub.com\u002Fdanielgatis\u002Frembg\u002Freleases\u002Fdownload\u002Fv0.0.0\u002Fsilueta.onnx)\n- [isnet](https:\u002F\u002Fgithub.com\u002Fxuebinqin\u002FDIS) | [isnet](https:\u002F\u002Fgithub.com\u002Fdanielgatis\u002Frembg\u002Freleases\u002Fdownload\u002Fv0.0.0\u002Fisnet-general-use.onnx)\n\n### TensorRT\n\nTensorRT 为 Nvidia GPU 提供优化的推理能力。模型必须转换为 TensorRT 引擎格式才能使用。这在受支持的硬件上能够提供最佳性能。\n\n## 故障排除\n\n有关故障排除的信息，请参阅 [故障排除文档](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fwiki\u002F06--Troubleshooting)。\n\n## 自行构建 ChaiNNer\n\n我在 GitHub 上提供了 ChaiNNer 的预编译版本。然而，如果您希望自行构建 ChaiNNer，只需运行 `npm install`（请确保已安装至少 npm v7）以安装所有 Node.js 依赖项，然后运行 `npm run make` 来构建应用程序。\n\n## 常见问题解答\n\n有关常见问题解答的信息，请参阅 [FAQ 文档](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fwiki\u002F07--FAQ)。\n\n## 文档\n\n如需深入了解 ChaiNNer 的各个方面，包括 CLI 使用、数据表示以及贡献者指南等，请参阅我们的 [ChaiNNer Wiki](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fwiki)。","# chaiNNer 快速上手指南\n\nchaiNNer 是一款基于节点流程图的图像处理 GUI 工具，最初专为 AI 图像超分辨率（放大）设计，现已发展为支持复杂图像处理流水线的强大平台。它允许用户通过拖拽和连接节点来定制处理流程，支持 Windows、macOS 和 Linux。\n\n## 环境准备\n\n### 系统要求\n- **Windows**: Windows 10 或更高版本（不支持 8.1 及以下）。\n- **macOS**: macOS 11 (Big Sur) 或更高版本（不支持 10.x 及以下）。\n  - *Apple Silicon (M1\u002FM2\u002FM3)*: 支持 PyTorch MPS 加速。\n- **Linux**: 主流发行版。\n  - *剪贴板功能*: 需安装 `xclip` 或 `wl-copy` (Wayland)。\n\n### 硬件与框架支持\nchaiNNer 启动时会自动下载独立的 Python 环境，**无需预先安装 Python**。根据显卡选择推荐的推理后端：\n- **NVIDIA 显卡**: 推荐 **PyTorch (CUDA)** 或 **TensorRT**（性能最佳）。\n- **AMD 显卡**: \n  - Linux: 推荐 **PyTorch (ROCm)**。\n  - 全平台: 可使用 **NCNN**。\n- **Intel 显卡**: 推荐使用 **NCNN**。\n- **CPU**: 所有框架均支持纯 CPU 模式作为备选。\n\n> **注意**: 若使用 NCNN 且遇到黑屏输出，请尝试在设置中手动调整分块（Tiling）大小。\n\n## 安装步骤\n\n### 1. 下载安装包\n访问 [GitHub Releases 页面](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Freleases\u002Flatest) 下载适合你系统的最新安装包（`.exe`, `.dmg`, 或 `.AppImage` 等）。\n\n*注：目前官方未提供国内镜像源，若下载速度慢，建议使用网络加速工具或寻找可靠的国内镜像站资源。*\n\n### 2. 运行安装\n- **Windows\u002FmacOS**: 双击运行下载的安装程序，按提示完成安装。\n- **Linux**: 赋予执行权限并运行（以 AppImage 为例）：\n  ```bash\n  chmod +x chaiNNer-*.AppImage\n  .\u002FchaiNNer-*.AppImage\n  ```\n\n### 3. 配置依赖\n首次启动时，chaiNNer 会自动下载并配置内置的 Python 环境。\n1. 点击界面右上角的 **Dependency Manager**（依赖管理器）按钮。\n2. 根据你的硬件选择并安装对应的神经网络框架：\n   - NVIDIA 用户：安装 **PyTorch** 或 **TensorRT**。\n   - AMD\u002FIntel 用户：安装 **NCNN** (或 Linux 下的 PyTorch ROCm)。\n3. 等待安装完成，其他 Python 依赖将自动处理。\n\n## 基本使用\n\n### 最简单的图像放大流程\n\n1. **添加节点**\n   - 在左侧面板找到节点，直接**拖拽**至编辑区，或双击节点名称。\n   - 依次添加以下三个节点：\n     1. `Load Image` (加载图片)\n     2. `Upscale Image` (放大图片)\n     3. `Save Image` (保存图片)\n\n2. **连接节点**\n   - 将鼠标悬停在节点边缘的连接点（Handle）上，拖动连线。\n   - 连接顺序：`Load Image` 输出 → `Upscale Image` 输入；`Upscale Image` 输出 → `Save Image` 输入。\n   - *提示：连接点颜色对应数据类型，系统仅允许兼容的节点相连。*\n\n3. **配置模型**\n   - 点击 `Upscale Image` 节点，在属性面板中选择已下载的超分模型（可从 [OpenModelDB](https:\u002F\u002Fopenmodeldb.info\u002F) 获取社区训练模型）。\n\n4. **运行流程**\n   - 点击顶部工具栏的绿色 **Run** 按钮。\n   - 观察连线动画：动画流动表示正在处理，动画停止表示该步骤完成。\n   - 处理完成后，检查输出文件夹即可看到放大后的图片。\n\n### 进阶技巧\n- **批量处理**: 使用 `Load Images` 节点代替 `Load Image` 可处理整个文件夹的图片。\n- **多选操作**: 按住 `Shift` 拖拽框选多个节点，按 `Delete` 键可批量删除。\n- **快捷菜单**: 在编辑区空白处右键，或从节点拉出连线后松开，可直接弹出兼容节点列表供快速选择。","一位独立游戏开发者需要将大量低分辨率的像素艺术素材批量放大并统一风格，以适配高清显示屏。\n\n### 没有 chaiNNer 时\n- **操作繁琐重复**：必须编写复杂的 Python 脚本或手动在 Photoshop 中逐个处理数百张图片，效率极低且容易出错。\n- **环境配置困难**：部署 AI 超分模型需要手动安装特定版本的 PyTorch、CUDA 驱动及各类依赖库，常因版本冲突导致运行失败。\n- **流程僵化单一**：难以将“放大”、“去噪”、“色彩校正”等多个步骤串联，每次调整参数都需重新运行整个脚本，缺乏灵活性。\n- **硬件适配麻烦**：在不同显卡（如 NVIDIA 与 AMD）之间切换时，往往需要重写代码或更换推理后端，学习成本高昂。\n\n### 使用 chaiNNer 后\n- **可视化流水线**：通过拖拽节点即可构建“加载图片→AI 超分→自动调色→保存”的自动化流程，一键批量处理所有素材。\n- **开箱即用体验**：chaiNNer 内置独立 Python 环境和依赖管理器，无需配置系统环境，启动即可直接调用 PyTorch 或 NCNN 等框架。\n- **高度灵活定制**：利用节点连线自由组合复杂任务，随时插入新的处理环节（如人脸修复），并实时预览中间结果以微调参数。\n- **跨平台兼容**：同一套节点流程图可在 Windows、macOS 和 Linux 间无缝迁移，自动适配不同用户的显卡硬件进行加速。\n\nchaiNNer 将原本高门槛的编程式图像处理转化为直观的可视化工作流，让创作者能专注于内容而非技术细节。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FchaiNNer-org_chaiNNer_a7c119be.png","chaiNNer-org","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002FchaiNNer-org_1a56c6df.png","Repositories related to the chaiNNer upscaling tool (WIP)",null,"chaiNNer.app","https:\u002F\u002Fgithub.com\u002FchaiNNer-org",[82,86,90,94,98,102],{"name":83,"color":84,"percentage":85},"Python","#3572A5",55.9,{"name":87,"color":88,"percentage":89},"TypeScript","#3178c6",42.7,{"name":91,"color":92,"percentage":93},"JavaScript","#f1e05a",0.8,{"name":95,"color":96,"percentage":97},"SCSS","#c6538c",0.6,{"name":99,"color":100,"percentage":101},"HTML","#e34c26",0,{"name":103,"color":104,"percentage":101},"Shell","#89e051",5710,347,"2026-04-05T01:52:54","GPL-3.0","Windows, macOS, Linux","非必需（支持 CPU 模式）。NVIDIA: 支持 CUDA (PyTorch), TensorRT; AMD (Linux): 支持 ROCm (PyTorch); 所有平台 AMD\u002FIntel: 支持 NCNN; Apple Silicon: 支持 MPS。具体显存和 CUDA 版本未说明，取决于所选后端和模型。","未说明",{"notes":113,"python":114,"dependencies":115},"1. macOS 10.x 及以下、Windows 8.1 及以下不支持。2. Apple Silicon Mac 上 ONNX 仅支持 CPU，NCNN 在某些配置下可能工作不正常。3. 部分非 NVIDIA 显卡用户使用 NCNN 可能出现黑屏输出（显存溢出），建议手动设置分块大小。4. Linux 用户使用剪贴板节点需安装 xclip 或 wl-copy。5. 首次运行会自动下载集成 Python 并通过依赖管理器安装其他库。","内置独立 Python 环境（推荐）；若使用系统 Python，需 3.10+ (推荐 3.11+)",[116,117,118,119,120],"PyTorch","NCNN","ONNX","TensorRT","Spandrel",[14,15],"2026-03-27T02:49:30.150509","2026-04-06T07:14:53.011687",[125,130,135,140,145,150],{"id":126,"question_zh":127,"answer_zh":128,"source_url":129},17923,"在 Arch Linux 上首次启动 chaiNNer 时遇到严重错误怎么办？","如果您通过 AUR 或旧版便携包安装遇到问题，请尝试直接从官网下载最新版本：https:\u002F\u002Fchainner.app\u002Fdownload。此外，请注意新版本已将捆绑的 Python 环境从 3.9 更新为 3.11，这通常能解决兼容性导致的启动崩溃问题。","https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fissues\u002F1809",{"id":131,"question_zh":132,"answer_zh":133,"source_url":134},17924,"为什么使用 CPU 进行超分辨率处理时会占用大量内存甚至导致内存泄漏？","这通常不是内存泄漏，而是预期行为。通过 CPU 进行超分辨率处理无法利用显存（VRAM），因此会消耗大量的系统 RAM。一旦处理完成，内存通常会被释放。如果处理大图像，高内存占用是正常的。","https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fissues\u002F783",{"id":136,"question_zh":137,"answer_zh":138,"source_url":139},17925,"运行 ONNX 模型时报错\"INVALID_ARGUMENT : Got invalid dimensions for input\"是什么原因？","此错误通常表示输入图像的通道数与模型期望的不匹配。例如，错误信息\"index: 1 Got: 1 Expected: 3\"表明模型需要 3 通道（RGB）图像，但实际输入的是单通道（灰度）图像。请确保在将图像送入模型前，使用节点将其转换为正确的通道格式（如 RGB）。","https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fissues\u002F808",{"id":141,"question_zh":142,"answer_zh":143,"source_url":144},17926,"在 macOS 上安装 PyTorch 依赖时提示找不到 torch==1.12.1 版本怎么办？","这是因为 PyPI 上该版本的 PyTorch 可能已不再支持当前的平台或已被归档。如果您需要使用特定版本（如为了 MPS 支持），可能需要手动指定索引源或接受更新的版本（如 2.0+）。如果是 facexlib 等依赖库报错，可能需要手动修改库文件以支持 MPS 设备，或者在该依赖库的仓库中提交 Issue。","https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fissues\u002F1891",{"id":146,"question_zh":147,"answer_zh":148,"source_url":149},17927,"MacOS Big Sur 上使用某些模型出现黑屏如何解决？","这可能是特定版本（如 0.11.5）在 MacOS 上的兼容性问题。维护者建议测试最新版本（如 0.14 或更高），因为许多针对 Windows 和 Mac 的底层图形队列（如 vkQueueSubmit）问题已在后续版本中修复。如果问题依旧，目前暂无针对 Mac 的注册表式修复方案，升级软件是主要解决途径。","https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fissues\u002F861",{"id":151,"question_zh":152,"answer_zh":153,"source_url":154},17928,"开发模式下运行 yarn dev 报错\"module '__main__' has no attribute '__file__'\"怎么处理？","该错误是由于后端框架 Sanic 更新了默认配置，从多线程（multithreading）改为了多进程（multiprocessing）模式，导致在某些导入场景下无法获取__file__属性。解决方法是在启动配置中显式将 worker 类型改回多线程，或者更新 chaiNNer 代码以适配新的 Sanic 多进程环境。","https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fissues\u002F1401",[156,161,166,171,176,181,186,191,196,201,206,211,216,221,226,231,236,241,246,251],{"id":157,"version":158,"summary_zh":159,"released_at":160},108225,"v0.25.1","这是一次小更新，主要修复了自上一版本发布以来发现的几个问题。如果您当前使用的 chaiNNer 版本运行正常，可以忽略本次更新。\n\n如果您使用的是较旧的（1000系列或更低）NVIDIA显卡，请通过依赖管理器重新安装PyTorch。\n\n## 变更内容\n* 修复GTX 1000系列显卡与PyTorch的兼容性问题（https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fpull\u002F3161）\n* 修复条件输入验证中的bug，该bug导致“非”逻辑运算需要隐藏输入B（https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fpull\u002F3154）\n* 修复保存视频节点中音频无法保存的问题（#3174）\n* 修复依赖项解析问题，防止使用全局Python site-packages（https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fpull\u002F3134）\n* 修复TensorRT引擎缓存跨应用重启后无法持久化的问题（https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fpull\u002F3139）\n* 修复Python安装错误信息中的语法和大小写问题（https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fpull\u002F3133）\n* 隐藏“安装所有包”按钮，因为其功能已损坏（https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fpull\u002F3179）\n\n注意：由于代码签名会导致构建问题，我不得不为macOS禁用代码签名。目前尚不清楚这会在尝试安装此更新时具体表现为何种情况，但特此提醒您，可能会出现以前未曾遇到的警告提示。\n\n**完整变更日志**：https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fcompare\u002Fv0.25.0...v0.25.1","2025-10-23T00:07:29",{"id":162,"version":163,"summary_zh":164,"released_at":165},108226,"v0.25.0","在超过一年没有稳定版本发布之后，chaiNNer 的开发工作已恢复，但目前仍以较低的产能进行。本次发布将夜间版中的功能引入到稳定版中，并修复了夜间版中报告的许多问题。\n\n主要的新功能是支持同时使用多个长度相等的迭代器。不过，嵌套或组合迭代仍然不被支持。根据用户反馈，该功能的用户体验仍较为有限，但已经趋于稳定。\n\n拥有 RTX 5000 系列显卡的用户应安装此更新，并通过依赖管理器更新 PyTorch。\n\n# 更改日志：\n\n## 主要变更：\n- 增加对组合迭代链的支持，由 @joeyballentine 和 @RunDevelopment 在 #2949、#2960、#2973、#2975、#2971、#3141 中实现。\n  - 这使得可以混合使用相同长度的迭代器。\n- 允许更多节点成为“起始节点”，由 @joeyballentine 在 #2944 中实现。\n  - 例如，现在可以将路径传递给“加载图像”节点。\n- 更新 Torch 和 Torchvision 以支持 RTX 50 系列，由 @LlamaEnjoyer 在 #3103 中完成。\n\n## 次要变更：\n- 为 PyTorch 上采样节点添加填充输入，由 @RunDevelopment 在 #2966 中实现。\n- 增加对三种 TIFF 压缩模式的支持，由 @RunDevelopment 在 #2970 中实现。\n- 添加 ONNX 的尺寸要求，并改进形状检测的鲁棒性，由 @RunDevelopment 在 #2978 和 #2979 中完成。\n- 根据 Spandrel 0.4.1 版本的支持情况，增加对新型超分辨率架构的支持，由 the-database 在 #3038 和 #3082 中实现。\n- 在“添加噪声”节点中增加对细微噪声生成的支持，由 @Arcitec 在 #3055 中实现。\n- 对“将图像对齐到参考图”节点进行更新，由 @pifroggi 在 #3091 中完成。\n- 将“模型缩放”节点更名为“获取模型信息”节点，由 @tumuyan 在 #3090 中完成。\n\n## 新增节点：\n- 增加“平衡法线”节点，由 @RunDevelopment 在 #2964 中实现。\n- 增加“逻辑运算”节点，由 @RunDevelopment 在 #2990 中实现。\n\n## 错误修复：\n- 修复撤销历史记录中包含选择状态的问题，由 @joeyballentine 在 #2961 中完成。\n- 修复比较节点中“!=”标签的问题，由 @RunDevelopment 在 #2991 中完成。\n- 澄清“加载模型”SafeTensors 文档，由 @JeremyRand 在 #3028 中完成。\n- （A1111\u002FSD）修复 SwinIR 4X 被错误命名引用的问题，由 @PereViader 在 #2998 中完成。\n- 修复依赖管理器中“安装所有包”选项无限挂起的问题，由 @copilot 在 #3127 中完成。\n\n## 已知问题：\n- “逻辑运算”节点目前在 NOT 运算上存在一个问题，正在进一步排查中。\n- ONNX 的 DML 支持由 @FNsi 添加，但目前仍存在故障。\n\n## 新贡献者\n* @PereViader 在 https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fpull\u002F2998 中完成了首次贡献。\n* @FNsi 在 https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fpull\u002F3072 中完成了首次贡献。\n* @Arcitec 在 https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fpull\u002F3055 中完成了首次贡献。\n* @styler00dollar 在 https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fpull\u002F3067 中完成了首次贡献。\n* @tumuyan 在 https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fpull\u002F3090 中完成了首次贡献。\n* @LlamaEnjoyer 在 https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fpull\u002F3103 中完成了首次贡献。\n* @oumad 在 https:","2025-10-19T22:07:30",{"id":167,"version":168,"summary_zh":169,"released_at":170},108227,"v0.24.1","本次更新包含若干 bug 修复和一些用户体验优化。\n\n## 变更内容\n* 由 @RunDevelopment 在 #2940 中添加锚点\u002F位置下拉菜单的 UI\n* 由 @RunDevelopment 在 #2931 中为依赖管理器添加“安装全部”按钮\n* 由 @RunDevelopment 在 #2938 中使节点宽度在处理手柄时更加一致\n* (PyTorch) 由 @joeyballentine 在 #2930 中添加每次推理后清除 CUDA 缓存的设置\n* (PyTorch) 由 @joeyballentine 在 #2943 中禁止自动更新 PyTorch，以确保自定义版本保持安装状态\n\n## Bug 修复\n* (PyTorch) 由 @joeyballentine 在 #2929 中修复超分优化后错误的瓦片尺寸估算\n* 由 @RunDevelopment 在 #2937 中修复分辨率节点的输出类型问题\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fcompare\u002Fv0.24.0...v0.24.1","2024-06-07T21:36:42",{"id":172,"version":173,"summary_zh":174,"released_at":175},108228,"v0.24.0","本次更新为 chaiNNer 带来了多项重大新功能，包括但不限于：条件分支（通过两个新节点——“Conditional”和“Compare”实现）、节点直通\u002F跳过、全新主题、PyTorch 优化，以及大量修复和体验改进。\n\n“Conditional”和“Compare”节点引入了一个备受期待的功能：分支支持。当然，目前还只是部分实现。这两个节点允许你根据对两个值进行不同方式比较后生成的条件，在多个输入之间切换。这意味着用户终于可以构建这样的工作流：在某一条件下执行一种操作，而在另一条件下执行另一种操作（例如，判断图像是否超过某个尺寸或低于某个尺寸）。需要注意的是，这种设置方式可能与直觉相反。由于是基于条件在不同值之间切换，决定的是条件_之前_的内容，而不是条件_之后_的内容。\n\n此外，“禁用”开关已被一个按钮取代，该按钮可在三种状态间切换：“启用”、“禁用”和“跳过”。在“跳过”模式下，节点不会被执行，但也不会像单纯禁用那样阻止下游节点运行。这也是用户呼声很高的功能之一，我们很高兴它终于加入到 chaiNNer 中了。\n\n现在可以在设置中选择新的主题。这一功能目前仍处于半实验阶段，因为我们计划在未来添加对完全自定义主题的支持，并进一步优化现有的默认主题选项。当前的主题选择主要是一个概念验证，旨在展示我们确实能够支持某种形式的主题，同时也为用户提供一些自定义选项。我们非常欢迎大家分享对现有主题的看法，以及对未来可添加主题的建议。\n\nPyTorch 超分辨率模型的性能得到了大幅优化。我使用了一款可视化分析工具（名为 VizTracer，强烈推荐给任何希望优化 Python 项目的人）来定位超分辨率处理中的性能瓶颈。经过优化后，超分辨率的速度大约提升了 25% 左右（当然，具体提升幅度会因所使用的模型及其他因素而有所不同）。虽然尚未达到完美，但这仍然是一次显著的改进，有助于加快 PyTorch 超分辨率工作流的运行速度。\n\n## 主要变更\n* 支持节点直通\u002F跳过功能，由 @RunDevelopment 在 #2805、#2864 和 #2887 中实现。\n* 新增主题功能，由 @joeyballentine 在 #2833、#2847、#2874 和 #2921 中实现。\n* 对 PyTorch 超分辨率进行了多项优化，由 @joeyballentine 在 #2853、#2852、#2861 等多个 PR 中完成。\n\n## 次要变更\n* 添加 AVIF 格式支持，由 @the-database 在 #2820 中实现。\n* 添加卷积用法示例，由 @BigBoyBarney 在 #2827 中实现。\n* （PyTorch）新增自定义分块选项，由 @the-database 在 #2828、#2870 和 #2880 中实现。\n* （PyTorch）更新 spandrel 库（新增对 PLKSR、RealPLKSR、IPT、SAFMN_BCIE 和 DRCT 的支持），由 @RunDevelopment 在 #2837 和 #2846 中完成。\n* 默认将链文件保存至“文档”文件夹，由 @joeyballentine 在 #2836 中实现。\n* 重命名数值输入参数并优化参数界面，由 @RunDevelopment 在 #2863 中完成。","2024-06-01T19:50:57",{"id":177,"version":178,"summary_zh":179,"released_at":180},108229,"v0.23.3","这是一个热修复版本，用于解决 v0.23.2 中引入的回归问题，该问题会导致迭代器在运行时出现错误。对此给受到影响的用户造成的不便，我们深表歉意。\n\n## Bug 修复\n* 由 @RunDevelopment 在 #2818 中修复了收集器的输入收集问题","2024-04-24T18:19:42",{"id":182,"version":183,"summary_zh":184,"released_at":185},108230,"v0.23.2","本次更新修复了若干 bug，并新增了一些功能。\n\n请注意：如果您在安装或更新 chaiNNer 后，设置完成时出现严重错误，请先尝试重启 chaiNNer，再进行报告。\n\n## 变更内容\n* 由 @RunDevelopment 在 #2795 中优化了不同 x-y 半径的方框模糊效果。\n* 由 @RunDevelopment 在 #2799 中将 FFMPEG 下载移至后台执行。\n* 由 @RunDevelopment 在 #2801 中改进了 `goIntoDirectory` 的验证逻辑。\n* 由 @RunDevelopment 在 #2800 中支持多节点的启用与禁用操作。\n* 由 @RunDevelopment 在 #2809 中对所有相对路径进行了验证。\n* 由 @RunDevelopment 在 #2808 中为“保存图像”功能新增了“跳过已有文件”选项。\n* 由 @RunDevelopment 在 #2812 中为阈值抗锯齿添加了平滑度参数。\n\n## 新增节点\n* 由 @RunDevelopment 在 #2794 中新增了“裁剪边框”节点，用于移除具有纯色边界的边框或画框。\n\n## Bug 修复\n* 由 @joeyballentine 在 #2815 中修复了在 Windows 系统上使用非英语语言时启动崩溃的问题。\n* 由 @RaySit 在 #2792 中防止了中键和右键点击触发标签页切换。\n* 由 @RunDevelopment 在 #2793 中修复了文件输入中的“复制文件名”选项。\n* 由 @RunDevelopment 在 #2797 中禁止对 Craft 模型进行 ONNX 转换。\n* 由 @RunDevelopment 在 #2806 中修复了 SSE 请求无法取消且导致 SSE 卡住的问题。\n\n## 新贡献者\n* @RaySit 在 #2792 中完成了首次贡献。\n\n**完整变更日志**：https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fcompare\u002Fv0.23.1...v0.23.2","2024-04-23T23:34:06",{"id":187,"version":188,"summary_zh":189,"released_at":190},108231,"v0.23.1","## 变更内容\n* 由 @RunDevelopment 在 https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fpull\u002F2785 中优化了 Clamp 节点\n* 由 @RunDevelopment 在 https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fpull\u002F2786 中将 Fast NL means 节点重命名为 Denoise\n\n## Bug 修复\n* 由 @joeyballentine 在 https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fpull\u002F2787 中修复了更改依赖项后节点不刷新的问题\n\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fcompare\u002Fv0.23.0...v0.23.1","2024-04-12T14:30:27",{"id":192,"version":193,"summary_zh":194,"released_at":195},108232,"v0.23.0","本次发布中最大的部分是我们进行的一系列内部改动，这些改动并不容易被用户直接感知到。不过，大家（希望）会在某些方面感受到更多的稳定性。\n\n此外，还包含了许多新功能、变更、错误修复以及一如既往的其他内容。\n\n和往常一样，请大家报告在本版本中遇到的新问题。\n\n## 主要变更\n* @joeyballentine 和 @RunDevelopment 在 #2597、#2647、#2649、#2651、#2650、#2659、#2675、#2677、#2683、#2718、#2719、#2726、#2738、#2745、#2748、#2754、#2755、#2767、#2769 中进行了大规模的内部重构：\n    * 提升了停止执行的可靠性\n    * 提升了从依赖管理器安装依赖的可靠性\n    * 使 UI 即使在使用远程主机时也能处理依赖安装（注：远程主机仍不完全支持）\n* @joeyballentine 在 #2749 中新增了通过双击节点名称来重命名节点的功能\n* @joeyballentine 在 #2731 和 #2758 中移除了启动画面，chaiNNer 现在会直接进入主窗口\n* @RunDevelopment 在 #2656 和 #2657 中为折叠节点添加了关键信息\n* （PyTorch）@RunDevelopment 在 #2705 和 #2732 中使 PyTorch 超分辨率过程可暂停\n* @RunDevelopment 在 #2768 中改进了 Python 的设置流程\n* @RunDevelopment 在 #2775 中改进了迭代器的预计剩余时间显示\n* @RunDevelopment 在 #2697 中允许在数字输入框中使用数学表达式\n\n## 依赖更新\n* （PyTorch）@RunDevelopment 和 @joeyballentine 在 #2655、#2710、#2661 和 #2646 中更新了 Spandrel，以提升对模型的支持：\n    * 新增对 RGT、Restormer、FFTformer、M3SNet、DCTLSA、APISR、MixDehazeNet、ATD、AdaCode、MPRNet、MIRNet2、DnCNN、FDnCNN 和 DRUNet 的支持\n    * 移除了对 SPSR 的支持\n* （ONNX）@joeyballentine 在 #2717 中更新了 ONNX 和 ONNX Runtime，修复了执行问题\n\n## 新节点\n* @RunDevelopment 在 #2695 中新增了“目录上一级”和“目录转文本”节点\n* @RunDevelopment 在 #2702 中新增了“进入目录”节点\n\n## 小幅变更\n* @RunDevelopment 在 #2640 和 #2735 中采用单一设置文件并重构了设置系统\n* @RunDevelopment 在 #2653 中使“随机种子”按钮不可拖动\n* @RunDevelopment 在 #2652 中新增了隐藏小地图的设置\n* @RunDevelopment 在 #2672 中在依赖管理器中使用软件包的显示名称\n* @RunDevelopment 在 #2685 中改进了错误报告机制\n* @RunDevelopment 在 #2690 中自动移除具有副作用的未使用节点\n* @RunDevelopment 在 #2691 中为“保存 X”节点自动创建目标目录\n* @RunDevelopment 在 #2692 中将图像类别中的重要节点移至列表上方\n* @RunDevelopment 在 #2698 中点击步进器后自动设置数值\n* @RunDevelopment 在 #2704 中忽略文档中的错误类型\n* @RunDevelopment 在 #2737 中改进了缺失模块错误的日志记录\n* @RunDevelopment 在 #2742 和 #2743 中改进了依赖管理器的界面\n* @RunDevelopment 在 #2746 中清理了卸载时的残留文件（仅限 Windows）\n* @RunDevelopment 在 #2746 中防止进度条遮挡 logo","2024-04-11T03:08:47",{"id":197,"version":198,"summary_zh":199,"released_at":200},108233,"v0.22.2","## 变更内容\n* @RunDevelopment 在 #2625、#2627、#2631 中进行了多项 UI 优化\n* @joeyballentine 在 #2582、#2637 中为编辑器节点选择面板添加了建议连接\n* @mrjschulte 将“Change Colorspace”节点重命名为“Change Color Model”，详见 #2574\n* @RunDevelopment 在 #2635 中将类型标签拉伸以填满可用空间\n* @RunDevelopment 在 #2638 中为“Create Noise”节点添加了平滑值噪声\n\n## 新增节点\n* @mrjschulte 在 #2610 中新增了“Resolutions”节点\n\n## 错误修复\n* @RunDevelopment 在 #2631、#2633、#2634、#2632 中修复了多项 UI 问题\n* @RunDevelopment 在 #2628 中修复了“Create Checkerboard”节点未能正确处理灰度和 RGB 图像的问题\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fcompare\u002Fv0.22.1...v0.22.2","2024-03-01T23:26:50",{"id":202,"version":203,"summary_zh":204,"released_at":205},108234,"v0.22.1","## 变更内容\n* 在 #2596、#2602 中由 @mrjschulte 添加 MiniMap\n* 在 #2613 中由 @RunDevelopment 支持 PyTorch 超分辨率的自定义缩放比例\n* 在 #2617 中由 @joeyballentine 对“保存视频”功能进行小幅优化\n\n## 新节点\n* 在 #2606、#2614 中由 @mrjschulte 和 @joeyballentine 添加“创建棋盘格”节点\n* 在 #2578 中由 @mrjschulte 添加“预乘 Alpha”节点\n\n## 错误修复\n* 在 #2619、#2618 中由 @joeyballentine 修复加载包含视频节点的某些旧版工作流的问题\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fcompare\u002Fv0.22.0...v0.22.1","2024-02-28T00:16:24",{"id":207,"version":208,"summary_zh":209,"released_at":210},108235,"v0.22.0","## Dependency Changes\r\n* Auto update dependencies by default, except GPU Pytorch & Torchvision by @joeyballentine in #2524\r\n    * Most dependencies will now be auto-updated if installed, to avoid any issues due to users not updating\r\n* (PyTorch) Update Spandrel to v0.2.2 by @joeyballentine in #2553\r\n    * Fixes SPAN model loading when created by updated neosr\r\n\r\n## Major Changes\r\n* Collapsing & expanding nodes by @joeyballentine, @RunDevelopment in #2544, #2560\r\n    * A new button on the top-left corner of nodes to collapse or expand nodes\r\n* Add \"Breakpoints\" feature to edges for custom edge routing by @joeyballentine, @RunDevelopment in #2548, #2583, #2586\r\n    * Right-click > Add Breakpoint or hold alt and click to add a \"breakpoint\" to a connection line\r\n* Arrow-key navigation in node search panel by @joeyballentine in #2575\r\n    * The viewport's right-click\u002Fdrag-out node search panel now supports arrow key navigation\r\n\r\n## Changes\r\n* Make iterator handles more square by @RunDevelopment in #2503\r\n* Generate tileable normal maps by @RunDevelopment in #2505\r\n* Make Fill Alpha output the alpha channel by @RunDevelopment in #2510\r\n* New inline label design for number, text, and dropdown inputs by @RunDevelopment in #2502\r\n* Change \"Note\" node styling and make it more versatile  by @joeyballentine in #2465\r\n* Remove the unfinished experimental presets feature by @joeyballentine in #2522\r\n* Update Load Video node for more accurate color conversion and chroma reconstruction by @mrjschulte in #2532\r\n* Add more inputs\u002Foutputs to Pass Through by @joeyballentine in #2549\r\n* Improve UI for Text as Image and Color Levels nodes by @RunDevelopment in #2552, #2554, #2561, #2564, #2571\r\n* Improve color accuracy in Remove Background node by @RunDevelopment in #2557\r\n* Hide Copy to Clipboard input label by @RunDevelopment in #2562\r\n* Improve UI of Normal Map Generator node by @RunDevelopment in #2567\r\n* Support hints for inline labels by @RunDevelopment in #2569\r\n* Improved node resetting by @joeyballentine, @mrjschulte in #2566, #2547\r\n\r\n## New Nodes\r\n* Add \"Accumulate\" node for sequences of numbers by @RunDevelopment in #2507\r\n* Add \"Edge Detection\" node by @RunDevelopment in #2509, #2590\r\n* Add \"Multiply\" node by @mrjschulte in #2540\r\n* Add \"Add\" node by @mrjschulte in #2528\r\n* Add \"Divide\" node by @mrjschulte in #2541\r\n* Add \"Log To Linear\" node by @mrjschulte in #2546\r\n* Add \"Create Colorwheel\" node by @mrjschulte in #2576\r\n* Add \"Clamp\" node by @mrjschulte in #2573\r\n\r\n## Bug Fixes\r\n* Fix load error messages not having a stack trace by @RunDevelopment in #2508\r\n* List images starting with dot by @RunDevelopment in #2511\r\n* Disallow ONNX conversions for SAFMN by @RunDevelopment in #2512\r\n* Save video input fix by @RunDevelopment in #2514\r\n* Move model to device after pytorch interpolation by @joeyballentine in #2525\r\n* Avoid undefined behavior in PyTorch by @RunDevelopment in #2570\r\n* Fix Threshold + AA resulting in wrong broadcasting by @RunDevelopment in #2592\r\n* Fix pixelate node resizing the image by @RunDevelopment in #2589\r\n* Yield in iterator to process events by @RunDevelopment in #2535\r\n* Quiet FFMPEG terminal output by @mrjschulte in #2543\r\n* Fix color of inline code in tooltips by @RunDevelopment in #2568\r\n\r\n## New Contributors\r\n* @mrjschulte made their first contribution in #2532\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fcompare\u002Fv0.21.2...v0.22.0","2024-02-20T00:15:44",{"id":212,"version":213,"summary_zh":214,"released_at":215},108236,"v0.21.2","This update fixes a few things as well as makes some general improvements. You know the drill.\r\n\r\n## Dependency Updates\r\n* (PyTorch) Update to spandrel v0.2.1 by @RunDevelopment in #2487\r\n  -   Adds support for SAFMN and fixes a few minor issues with upscaling small images.\r\n\r\n## What's Changed\r\n* Add \"Copy\" to right-click context menus by @joeyballentine in #2473\r\n* Make Save Video input names consistent with Save Image by @RunDevelopment in #2480\r\n* Correctly show mixed iterator-non-iterator inputs\u002Foutputs by @RunDevelopment in #2481\r\n* Avoid running non-chosen \"Switch\" node paths by @RunDevelopment in #2477\r\n* Pad images according to size requirements when upscaling by @RunDevelopment in #2485\r\n* Add PyTorch memory budget limit setting by @joeyballentine in #2472\r\n* Add \"Manual\u002FCopy\" install mode to dependency manager by @joeyballentine in #2462\r\n* Mark iterated inputs\u002Foutputs as sequenced in docs by @RunDevelopment in #2495\r\n* Fix iterator limit and improved output types by @RunDevelopment in #2496\r\n* Fix iterator lineage checks in UI by @RunDevelopment in #2494, #2498\r\n* Allow handles for \"Merge Spritesheet\" number inputs by @RunDevelopment in #2499\r\n* Rename \"Strengthen Normals\" to \"Scale Normals\" by @RunDevelopment in #2500\r\n\r\n## New Nodes\r\n* \"Execution Number\" node by @joeyballentine in #2433\r\n\r\n## Bug Fixes\r\n* Fix scale of Hq2X in output type by @RunDevelopment in #2467\r\n* Fix NCNN model interpolation execution context by @RunDevelopment in #2468\r\n* Fix accent color for `any` outputs by @RunDevelopment in #2479\r\n\r\n## Other\u002FDev Changes\r\n* ONNX upscale reworked by @adegerard in #2417\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fcompare\u002Fv0.21.1...v0.21.2","2024-01-24T03:47:42",{"id":217,"version":218,"summary_zh":219,"released_at":220},108237,"v0.21.1","This update fixes a pretty crucial bug with upscaling images with transparency with PyTorch models.\r\n\r\nImportant note: I forgot to mention last update to make sure you have the latest Nvidia drivers after you update PyTorch. If your drivers haven't been updated in over a year, PyTorch upscaling might start not working or being extremely slow.\r\n\r\n## What's Changed\r\n* Better node errors using error messages in output types by @RunDevelopment in #2445\r\n* Add a popup message on start for dependency updates by @joeyballentine in #2453, #2458\r\n\r\n## Bug Fixes\r\n* Fixed upscaling with transparency by @RunDevelopment in #2457\r\n\r\n## Other\u002FDev Changes\r\n* Replace global execution options with node context parameter by @RunDevelopment in #2444\r\n* Use screenToFlowPosition() instead of project() by @joeyballentine in #2459\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002FchaiNNer\u002Fcompare\u002Fv0.21.0...v0.21.1","2024-01-13T17:05:15",{"id":222,"version":223,"summary_zh":224,"released_at":225},108238,"v0.21.0","This might be chaiNNer's biggest update ever! Sorry for how long it took, we spent a lot of time making sure this release had the polish we deemed necessary. There are a lot of big changes in this update, but here are some highlights: Better iterators, general optimizations, support for more PyTorch models, chain auto-organization, and much more. Here are the major changes in greater detail:\r\n\r\n### Better Iterators\r\nThis might not be the final step in our goal of getting iteration in chaiNNer to where we want to be, but it's certainly a step in the right direction. Iterators are no longer gigantic nodes with their own sub-flow editor in them. Now, they are single nodes that can be attached like normal, which makes working with iteration a whole lot easier. Not only that, but it also allows mixing and matching iterators and \"collectors\", so now you can do things like convert an image sequence to a video, or split a spritesheet to separate images.\r\n\r\nHowever, only one iterator is actually allowed to be part of the same chain \"lineage\", so you still are not able to do things like match multiple iterators together or combine an image iterator and a model iterator. That is a lot more complex and will hopefully come in the future. Oh yeah, and all the iterator nodes have new names as well, so instead of \"Image File Iterator\", look for \"Load Images\".\r\n\r\n### General Optimizations\r\nWe spent some time working on optimizing a few aspects of chaiNNer and generally improving speed. From making PyTorch upscales faster to optimizing things like resizing to making the frontend use less CPU, you should notice a general speedup all around.\r\n\r\n### Support for more PyTorch models\r\nPart of what made this release take a while was that we took a small break from working on chaiNNer to separate out our model support code into a new Python package called [Spandrel](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002Fspandrel). Doing so allowed us to more easily add support for a variety of new models (such as SPAN, Real-CUGAN, FBCNN, and more) and we even got some contributions from the community. Spandrel is now being used in AUTOMATIC1111's stable diffusion webui, and will most likely also be used in ComfyUI soon. If you want chaiNNer's automatic model support in your Python project, go check out Spandrel. To see a list of all the currently supported models, check [here](https:\u002F\u002Fgithub.com\u002FchaiNNer-org\u002Fspandrel\u002Ftree\u002Fv0.1.7?tab=readme-ov-file#model-architecture-support).\r\n\r\n### Other notable changes\r\n- Chains can now be auto-formatted\u002Forganized using Edit > Format Chain (or using ctrl+shift+f).\r\n- Upscale tiling is now even more seamless, using a blending technique to avoid hard seams with certain models like SCUNet.\r\n- Some general UI improvements and quality-of-life additions.\r\n- Many new nodes.\r\n\r\nAnd now for the changelog:\r\n\r\n## Dependency Updates\r\n\r\n### PyTorch\r\n* Update PyTorch to 2.1.2 by @joeyballentine in #2265, #2349, #2429\r\n* Add SafeTensors support by @joeyballentine in #2272, #2440\r\n* Use new Spandrel package for model loading by @joeyballentine and @RunDevelopment in #2330, #2343, #2348, #2354, #2359, #2392, #2412, #2428, #2409\r\n\r\n## Changes\r\n\r\n### New Nodes\r\n* \"Alpha Matting\" node by @RunDevelopment in #2292\r\n* \"Pixel Art Upscale\" node by @RunDevelopment in #2326\r\n* \"Strengthen Normals\" node by @RunDevelopment in #2372\r\n* \"Unified Pad\" node by @RunDevelopment in #2373\r\n* \"Unified Resize\" node by @RunDevelopment in #2380\r\n* \"Chroma Key\" node by @RunDevelopment in #2381\r\n* \"Optimize ONNX Model\" node by @RunDevelopment in #2408\r\n\r\n### Optimizations\r\n* Optimize box blur by @RunDevelopment in #2325\r\n* Optimize gamma implementation by @RunDevelopment in #2386\r\n* Optimize image resizing and add more resizing algorithms by @RunDevelopment in #2387, #2390, #2394\r\n* Optimize Split Transparency output for subsequent operations by @RunDevelopment in #2391\r\n* Optimize PyTorch upscale by @RunDevelopment in #2407\r\n* Optimize regex replace by @RunDevelopment in #2411\r\n* Change edge running animation and behavior by @joeyballentine in #2424\r\n\r\n### Changes\r\n* Iterator rewrite by @joeyballentine and @RunDevelopment in #2254, #2267, #2276, #2280, #2286, #2442, #2449\r\n* Add chain auto-organization\u002Fformatting by @joeyballentine and @RunDevelopment in #2377, #2395\r\n* Add type tags for number ranges by @RunDevelopment in #2284\r\n* Improve \"Remove Background\" node types by @RunDevelopment in #2291\r\n* Simple fused input-output elements by @RunDevelopment in #2338\r\n* Improve screenshots by @RunDevelopment in #2340\r\n* Add fused output for \"Pass Through\" node by @RunDevelopment in #2341\r\n* Allow configuring multithreading for ncnn by @JeremyRand in #2342\r\n* Make ncnn memory budget configurable by @JeremyRand in #2351\r\n* Allow setting ONNX opset by @RunDevelopment in #2361\r\n* Add overlap blending for auto split by @RunDevelopment in #2363\r\n* Add open delay to status tooltips by @RunDevelopment in #2364\r\n* Add threshold group by @RunDevelopment in #2368\r\n* Add \"tabs\"","2024-01-12T01:54:10",{"id":227,"version":228,"summary_zh":229,"released_at":230},108239,"v0.20.2","Hey everyone, sorry it took so long to get this out. This update contains a bunch of things that I thought we had already released -- a couple bug fixes and some cool new features. Some very big things are on the way though, so be on the lookout for the next update.\r\n\r\n## Changes\r\n- Add Multi Gaussian filter to Normal Map Generator (#2255)\r\n- Add ability to limit the length of certain iterators (#2218)\r\n- Add support for sub-pixel distance transform (#2235) \r\n- (ncnn) Allow disabling Winograd\u002FSGEMM (for cpu ncnn) (#2237 -- thanks @JeremyRand) \r\n\r\n## Bug Fixes\r\n- Fixed being able to close the dependency manager while installing deps (#2227)\r\n- Fixed RGBA preview image background (#2225) \r\n- Fixed pip using bad caches on dep installs (#2232)\r\n- Fixed dragging bug on image previews (#2236)\r\n- Fixed bug from read-only image in CAS (#2239)\r\n- Fixed CAS for grayscale images (#2241)\r\n\r\nAnd thanks to these maintainers: @joeyballentine and @RunDevelopment ","2023-10-14T04:44:17",{"id":232,"version":233,"summary_zh":234,"released_at":235},108240,"v0.20.1","This is a hotfix to fix the Windows installer installing to the wrong place. If you installed v0.20.0 on Windows, I recommend checking if your system registers two separate chaiNNer installs, and uninstalling the extra one after applying this update. If your system is not registering a second chaiNNer instance, I recommend navigating to `C:\\Users\\[your user]\\AppData\\Local` and seeing if a directory called `app` exists, and if it contains `chaiNNer.exe`. If it does, I recommend just deleting that `app` directory entirely. The v0.20.1 installer ***will not*** automatically fix this for you. I am very sorry for this inconvenience. \r\n\r\n## Bug Fixes\r\n- Fix windows installer installing to the wrong place (#2219)\r\n- Fix PyTorch erroring when both CPU and FP16 modes were enabled (#2217)","2023-09-16T20:57:09",{"id":237,"version":238,"summary_zh":239,"released_at":240},108241,"v0.20.0","Hey everyone! This is a pretty major update with some pretty overdue changes. The main big change is that we've upgraded to Python 3.11 (for integrated Python) as well as PyTorch 2.0 (if you're using chaiNNer for its PyTorch capabilities). Between these two changes, you should hopefully see a bit of an overall speed boost when processing things. That isn't all that got changed though, so take a look at the rest of the changelog to see everything that's new.\r\n\r\nNOTE: After installing v0.20.0, integrated Python will be redownloaded, and you will need to re-install any dependencies you are using from the dependency manager (assuming you are not using system Python). Also, dependency-related settings have been reset in this build, so make sure to double-check that everything looks correct)\r\n\r\n## Major Changes\r\n- Update Integrated Python to Python 3.11 (#2144, #2092)\r\n- _View Image_ nodes now have resizable image previews (#2167, #2174, #2205, #2214)\r\n- Improved _Video Frame Iterator_ codec and container selection (#2158 -- thanks @adegerard)\r\n- Add ability to pass custom additional ffmpeg params to _Video Frame Iterator_ (#2158 -- thanks @adegerard)\r\n- MacOS builds are now code-signed (#2136, #2180 -- thanks @stonerl)\r\n- MacOS builds are now universal (x64 and arm64) (#2162, #2166 -- thanks @stonerl)\r\n\r\n## Minor Changes\r\n- (PyTorch) DAT and SRFormer support (#2153, #2155, #2156)\r\n- (PyTorch) Support for reading .ckpt files (for certain models) (#2193)\r\n- (PyTorch) Support for DiffBR's SwinIR models (#2202)\r\n- (PyTorch) FP16 now gets enabled by default for compatible GPUs (#2203)\r\n- Linux build improvements (#2179 -- thanks @stonerl)\r\n- Wayland Linux support (#2178 -- thanks @illode)\r\n- Added BC5_SNORM format and additional format information to DDS options in _Save Image_ (#2199)\r\n- Added support for uncompressed DDS (#2201, #2211)\r\n- Performance improvement for iterator progress updates that should use less CPU and lag the UI less (#2182)\r\n- Visually show feature-disabled nodes as disabled (#2194)\r\n- Added docs for Mip options in _Save Image_ (#2207)\r\n- Added anti-aliasing to Threshold node (#2209)\r\n\r\n## New Nodes\r\n- _Text as Image_ (#2125, #2213 -- thanks @adegerard)\r\n    - This node allows you to create an image from text, for use with overlaying on other images.\r\n\r\n## Dependency Updates\r\n- Update PyTorch to 2.0.1 (#2143)\r\n- Update ONNX-related dependencies for better 3.11 support (#2153)\r\n\r\n## Bug Fixes\r\n- Fix .chn files being unable to be opened with chaiNNer on MacOS (#2150 -- thanks @stonerl)\r\n- Fix BC5 DDS loading (#2198)\r\n- Fix High Pass for grayscale images (#2206)\r\n- Fixed logging empty lines (#2208)\r\n\r\nand thanks to the following maintainers: @joeyballentine @rundevelopment @stonerl","2023-09-16T01:39:10",{"id":242,"version":243,"summary_zh":244,"released_at":245},108242,"v0.19.4","This update contains a few bug fixes, a few new features, and a bunch of improvements specifically for macOS.\r\n\r\n## Changes\r\n- Add position to _Blend Images_ node (#2087 -- thanks @adegerard)\r\n- Made some node outputs more consistent (#2111 -- thanks @stonerl)\r\n- Add \"Separate Alpha\" checkbox to upscale nodes (#2127)\r\n- Handle unsaved changes on chaiNNer restart (#2132 -- thanks @stonerl)\r\n- Various macOS DMG improvements (#2122 -- thanks @stonerl)\r\n- Set minimum macOS version to 11.0.0 (#2124 -- thanks @stonerl)\r\n- Remove the macOS portable build as it was broken and has likely never worked (#2119 -- thanks @stonerl)\r\n- Associate .chn files with chaiNNer on macOS (#2115 -- thanks @stonerl)\r\n- Add \"New Chain\" and \"Open Recent\" menus to macOS dock (#2140 -- thanks @stonerl)\r\n- Slight optimization for NCNN model loading when using CPU NCNN (#2142 -- thanks @JeremyRand)\r\n\r\n## Bug Fixes\r\n- Fixed bug related to opening certain older chain files (#2116)\r\n- Fixed tooltips being placed behind modal popups (#2128 -- thanks @stonerl)\r\n- Fixed hotkeys (such as pressing F5 to run a chain) sometimes not working (#2114 -- thanks @stonerl)\r\n- Fixed incorrect inputs on new unified _Crop_ node (#2134 -- thanks @adegerard)\r\n\r\nAnd thanks to the following maintainers: @joeyballentine @RunDevelopment","2023-08-25T22:02:30",{"id":247,"version":248,"summary_zh":249,"released_at":250},108243,"v0.19.3","Another update which adds a few things and fixes a few bugs.\r\n\r\n## Changes\r\n- (PyTorch) SCUNet architecture support (#2102, thanks @adegerard) \r\n- Threshold node improvements (#2091)\r\n\r\n## New Nodes\r\n- _Generate Threshold_ (#2096)\r\n    - A node to generate the threshold value for the Threshold node. \r\n\r\n## Bug Fixes\r\n- Fixed _Save Image_ not creating directories correctly (#2099)\r\n- Fixed issue with displaying image previews (#2101)\r\n- Disable multi-instance support on MacOS as it did not work correctly (#2107, thanks @stonerl)\r\n- Display modal overlays on top of toast messages (#2110, #2112, thanks @stonerl)\r\n\r\nAnd thanks to the following maintainers: @rundevelopment","2023-08-18T02:14:16",{"id":252,"version":253,"summary_zh":254,"released_at":255},108244,"v0.19.2","This update contains a few important bug fixes as well as some new long-requested features.\r\n\r\n## Changes\r\n- Use Rust implementation for dither nodes (#2075)\r\n    - This implementation is between 100 and 500 times faster than the previous pure-python implementation.\r\n- Use Rust implementation for _Copy to Clipboard_ (#2079)\r\n    - This removes the multiple platform-specific dependencies as well as enables support for this node on silicon Macs.\r\n- Add support for saving 16 bit and 32 bit images (#2083) \r\n- Add setting for allowing multiple concurrent chaiNNer instances (#2089)\r\n- Light mode improvements (#2064, thanks @stonerl)\r\n- Add support for saving BMP in _Save Image_ (#2077)\r\n- \"Webp (Lossless)\" in _Save Image_ is now just \"Webp\" with a lossless checkbox (#2080)\r\n- Added warnings in node documentation for nodes with limited color depth (#2090) \r\n\r\n## New Nodes\r\n- _Quantize to Reference_ (#2071)\r\n    - A custom quantization implementation that tries to emulate a reference image as closely as possible. Useful for pixel art upscales.\r\n- _Percent_ (#2076, thanks @adegerard)\r\n    - Similar to the _Number_ node, this node lets you input a percentage via slider to connect to other nodes.\r\n\r\n## Bug Fixes\r\n- Fixed dropdown settings resetting their values unintentionally (#2081)\r\n    - This fixes the ONNX execution option problem people have been reporting recently. Make sure to check that you aren't using CPU by accident.\r\n- Fixed _Image File Iterator_ not sorting alphabetically (#2067)\r\n- Fixed _Stretch Contrast_ (#2082)\r\n- Ignore file extension case in file image iterator (#2084)\r\n\r\nThanks to the following maintainers: @joeyballentine @RunDevelopment @theflyingzamboni","2023-08-14T21:40:40"]