V13.0 🚨 THE LEGACY COLLECTION (50+ Workflows) - FINAL UPLOAD 🚨
EN: This is the complete archive of my collection containing over 50 workflows. I have updated many of them one last time to ensure compatibility with the current version of Comfy UI. This is the Final Upload for this specific page.
I am making a clean cut: From January 2026 onwards, all my workflows will exclusively use my new, organized layout. Instead of one massive, unmanageable collection, I will release specialized packages (e.g., Nunchaku Flux, Z-Image, SDXL, Pony, etc.).
Why? This allows me to guarantee quality and keep everything up-to-date without losing track of 50+ individual files.
Thank you so much for your support over the years! Make sure to visit me on YouTube to stay tuned for the new era of workflows. This page will no longer be maintained.
DE: Hier ist das komplette Archiv meiner Sammlung von über 50 Workflows. Ich habe viele davon ein letztes Mal überarbeitet, damit sie mit der aktuellen Comfy UI Version kompatibel sind. Dies ist der letzte Upload (Final Upload) für diese Seite.
Ich mache jetzt einen klaren Schnitt: Ab Januar 2026 werden alle meine Workflows ausschließlich im neuen, aufgeräumten Layout erscheinen. Anstatt einer riesigen, unhandlichen Sammlung werde ich spezialisierte Pakete veröffentlichen (z. B. Nunchaku Flux, Z-Image, SDXL, Pony usw.).
Warum? So kann ich Qualität und Aktualität garantieren, ohne bei über 50 einzelnen Dateien den Überblick zu verlieren.
Vielen Dank für euren jahrelangen Support! Besucht mich unbedingt auf YouTube, um den Start der neuen Workflow-Ära nicht zu verpassen. Diese Seite wird ab sofort nicht mehr gepflegt.V12 Our Comfy UI workflows are updated to the latest version, enhanced with powerful tools like SDXL, Flux, Nuchaku, Kolors, Ollama, F5TTS, and WAN2.2 for advanced image, video, and multimedia creation.
To use Nuchaku, you need an RTX graphics card, and you'll need to do a few things in addition to the custom node installation. This is all explained in the four videos. A few days ago, a completely updated installation guide was also released for the model creators. The other workflows should all work simply by "Install missing custom nodes."
V11 New Nunchaku Workflows (v6) with Flux Kontext
V10 New Nunchaku Workflows (v5) with PuLID
V9.0 New Nunchaku Workflows (v4)
V8.01 Important information: New Wheel -> https://www.youtube.com/channel/UCX3CsqqI_MtPiNPreR16q7Q/community?lb=UgkxwDjR9B3P6XOsf_Sp6oCl98dZ68HBT8LF
V8.0 ✨Nunchaku Inpaint Speed Masterclass in ComfyUI: Top Quality, Blazing Fast! 🚀
New Nunchaku workflows adedd
This powerful workflow leverages Nunchaku for a wide range of creative tasks: from Text to Image (T2I) ✍️➡️🖼️ and Image to Image (I2I) 🖼️➡️🖼️ generation to precise Inpainting 🪄 and even a combination of these modes. It also integrates advanced features like Redux, Loras, Florence2, Murphy's Prompt Mixer 🧪, and many other fine-tunings ⚙️ to help you generate truly perfect images. 💎
V7.0 🚀Nunchaku + Comfy extremely fast! ✨
New Nunchaku workflows and 4k upscale workflow added
V6.5✨Restore 📸 Old Photos Like Magic!🪄
New Supir workflows added, minor bug fixes, image comparison workflow added.
V6.0 🚀 Features Optimizations different workflows for the latest ComfyUI version!
ComfyUI & Redux: Magical Image Generation Explained! 🎨✨ | AI Art with Flux & Redux
V5.0 🚀 Optimizations across almost all workflows for the latest ComfyUI version!
🚀 "Ollama Meets ComfyUI: Easy AI Workflows for Beginners & Pros!" 🤖
🔥 Florence 2 in ComfyUI – Next-Level AI Masking & Prompt Magic! 🎨✨
V4.01 🚀 Minor optimizations across almost all workflows – Improved performance
V4.0 🚀 Major Update: Murphy's Ultimate Workflow Collection for the New Comfy UI!
V3.01 Easy play with the multi workflows
V3.0 new easy maschine and multi maschine update, and and many small simple workflows
V2.5 new multi workflows
V2.22 - 15 new Easy workflows in the package
Introducing the must-have AI Image Upscaling Workflow, your secret weapon for transforming ordinary images into extraordinary masterpieces. Packed with the power of SDXL and Flux, this innovative workflow will guide you step-by-step through the process of enhancing even the smallest details with precision and finesse. Say goodbye to blurry faces and hello to professional-quality results, all at the click of a button. Whether you're a seasoned AI artist or just starting out, this workflow is a game-changer you simply can't afford to miss!
Face Crop, Inpaint Optimize, and various Detailer options provide precise image manipulation
Face Swap, Upscaler, SD Ultimate, Supir, Loras, and ControlNet offer ultimate creative control
Image 2 Image allows for stunning image transformations
Stylers, Flux, and Kolors enable you to add unique artistic flair
SDXL ensures lightning-fast, high-quality upscaling
Integration of multiple samplers encourages versatile experimentation
Optimization of existing images guarantees top-notch quality
The Comfy Workflow offers a comprehensive set of tools and features to enhance your image generation experience. With the ability to create, modify, and optimize images using powerful samplers like SDXL, Kolors, and Flux, you have unparalleled control over your creative process. The workflow also includes advanced features such as Controlnet, various optimization tools, and the ability to upscale your images with ease.
Some key capabilities of the Comfy Workflow include:
Generating high-quality images using SDXL, Kolors, and Flux samplers
Employing Controlnet for enhanced image control and precision
Optimizing and refining images with powerful tools
Upscaling images with the Ultimate SD Upscaler
Saving and organizing images with customizable settings
The Comfy Workflow is designed to be user-friendly and adaptable to your needs, making it the ultimate tool for artists, designers, and anyone looking to push the boundaries of digital image creation.
Description
New Flux context workflows added, all workflows revised
FAQ
Comments (11)
Excellent work! Your efforts made a real difference.
Wow, thank you very much for your comment, I'm glad you like my work.
I've never thanked you before, but I will now, thanks for all your meticulous work, you're the final boss of workflow creators!
Thank you so much, I'm very pleased. I still have some crazy ideas in my head of what else I could do, but there just isn't enough time besides my regular work.
Failed to find the following ComfyRegistry list. The cache may be outdated, or the nodes may have been removed from ComfyRegistry.
comfyui-nunchaku
你好,我现在遇见了这个问题,我不知道该怎么解决。
之前你推荐我过几天看教程,但是由于语言的差异,我不知道应该看哪一个教程。
能否指点一些我想走你的工作流的学习路线。
我目前是想要通过深度或线图控制生成结果,我应该选择你的那个工作流。
comfyui-nunchaku的问题是需要下载一个工作流并运行一次就不会报错了。
工具工作流程 — ComfyUI-nunchaku 0.3.3 文档
这是工作流下载位置。
duoye
Den Fehler kann ich nicht nachvolziehen, ich habe gerade ComfyUI auf die neuste Version aktuallisiert, da funktioniert alles. Sie wissen das sie eine NVIDIA Grafikkarte benötigen. Die Beispiel funktionieren für die 30er und 40er karten für 20er oder 50er Karten müssen sie Anpassungen die auf der Nunchaku Hompage beschrieben sind anwenden. https://github.com/nunchaku-tech/ComfyUI-nunchaku
Wenn es um die Logik der Workflows geht, das war das Video das ich gemeint habe:
5.) https://youtu.be/ZBVv6iEttrw
Ich weiß leider nicht was sie mir sagen wollen. Wenn es um Nunchaku geht würde ich folgende Reihenfolge empfehlen. Allerdings hat sich etliches verbessert und manches aus den "Ersten Videos ist schon überholt, aber wenn sie sich alle ansehen sollten sie verstehen was sich vereinfacht hat, bei der Installation. Die Systematik bei den Workflows funktioniert aber durchgänig gleich. Die neusten Videos sollten eine automatische Übersetztung haben (Sprache und Untertitel) wie gut die ist kann ich nicht beurteilen.
1.) https://youtu.be/rtWyD2kf9cI
2.) https://youtu.be/Z5vWEwa0cEY
3.) https://youtu.be/7vMj49DuTgY
4.) https://youtu.be/HAq2uXbr_h0
Alle Videos 2025 unter: https://www.youtube.com/playlist?list=PLOlk3DTLbLzSJzShrwN0m9GWrlA8BxDU1
翻译过了,但我不知道翻译得如何:
我无法重现错误。我刚刚将 ComfyUI 更新到最新版本,一切正常。您知道您需要一块 NVIDIA 显卡。这些示例适用于 30 和 40 系列显卡。对于 20 或 50 系列显卡,您需要按照 Nunchaku 主页上的说明进行调整。https://github.com/nunchaku-tech/ComfyUI-nunchaku
至于工作流程逻辑,我指的是这个视频:
5.) https://youtu.be/ZBVv6iEttrw
很遗憾,我不知道您想告诉我什么。关于 Nunchaku,我建议按照以下顺序操作。虽然现在有很多改进,而且一些“最初”的视频已经过时了,但如果您全部看完,您应该就能理解在安装过程中哪些地方被简化了。不过,工作流程系统始终保持不变。最新的视频应该有自动翻译(语言和字幕),但我无法判断它有多好。
1.) https://youtu.be/rtWyD2kf9cI
2.) https://youtu.be/Z5vWEwa0cEY
3.) https://youtu.be/7vMj49DuTgY
4.) https://youtu.be/HAq2uXbr_h0
所有 2025 个视频位于:https://www.youtube.com/playlist?list=PLOlk3DTLbLzSJzShrwN0m9GWrlA8BxDU1
duoye 是的,确切要下载轮子,您必须运行此工作流1倍。如果您观看视频,则应该了解它的意思。
解决了comfyui-nunchaku这个问题,工作流可以直接跑,但是我是20系显卡,前几次跑闪退了,问AI有提到20系显卡会概率出现闪退问题,所以非30,50系显卡遇见类似的问题可以尝试多跑几次看看是否能解决问题。
不过请注意30系和50系可能要选择的模型不同,请留意备注。
murphylanga 十分感谢你远方的朋友,愿意回复我这么多,甚至专门为我做了翻译,让我感到受宠若惊。
现在我的问题已经彻底解决,不单纯是工作流上的问题,我的显卡是2060所以这里存在一些复杂的问题。
我尝试了很多办法,有你提供的一些信息,现在终于得到了解决。
我也看过官方3.3的文档但是我没有得到能解决我问题的办法。
我准备明天找个时间做个详细的教程放在你的评论区希望能帮助到其它人显卡成功部署你的工作流。
至于现在我人已经热傻了,我只想学一只猴子开始欢呼。
(其实给我的回复不必刻意翻译,我的浏览器有内置的翻译,只是在youtube不是很方便)
我20系显卡跑通这个工作流的过程十分痛苦,所以我想你们会需要一份详细的教程,下面是我的详细安装过程。
在以下网站获取预编译文件
https://github.com/nunchaku-tech/nunchaku/releases?page=2
例如
nunchaku-0.3.1+torch2.5-cp310-cp310-win_amd64.whl
然后cd到下载路径进行安装
python -m pip install nunchaku-0.3.1+torch2.5-cp310-cp310-win_amd64.whl
重装PyTorch 2.5 + CUDA 版卸载
pip uninstall torch torchvision torchaudio -y
安装CUDA
pip install torch==2.5.1+cu121 --index-url https://download.pytorch.org/whl/cu121
安装onnxruntime加速
pip install onnxruntime-gpu
安装torchvision
pip install torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
修改下载的nunchaku文件版本为预编译版本,无视弹出的报错
找到这个节点
Nunchaku FLUX DiT Loader Flux Depth
data_type的默认选项我记得是bfloat16,这个20系显卡不支持要点击换成float16
同样的20系的问题不支持awq-int4-flux.1-t5xxl.safetensors模型
下t5xxl_fp8_e4m3fn_scaled.safetensors模型代替
模型下载位置
CLIP1
ViT-L-14-TEXT-detail-improved-hiT-GmP-HF
CLIP2 30/40系显卡
awq-int4-flux.1-t5xxl.safetensors
https://huggingface.co/nunchaku-tech/nunchaku-t5/tree/main
CLIP2 20系显卡
t5xxl_fp8_e4m3fn_scaled.safetensors
https://huggingface.co/comfyanonymous/flux_text_encoders/tree/main
VAE 你也可以尝试其它的
Flux vae.sft 你
https://civitai.com/models/636193/flux-vaesft?modelVersionId=711305
flux.1模型(50系下载fp4)
https://huggingface.co/mit-han-lab/nunchaku-flux.1-depth-dev
注:
我自己跑的工作流是Nunchaku文件下的Canny_Multi_Flux_Nunchaku_Diffusion_Comfort_Combine_Workflow_v6.0
如果需要更详细的使用教程,这个工作流的详细页面有每个工作流的介绍,不过我会直接推荐你看
V11 New Nunchaku Workflows (v6) with Flux Kontext
这个教程,并使用Nunchaku文件下的工作流。

