Share the best and most practical ComfyUI workflow, the latest and most cutting-edge research results and corresponding custom nodes, practical videos, text tutorials and knowledge, etc.
QQ communication group: 801334535.
| Gallery (reference) | Official, GitHub | High quality video | resource |
|---|---|---|---|
| ideogram | ComfyUI | Nerdy Rodent | OpenArt |
| midjourney-showcase | ComfyUI-Manager | Latent Vision | civitai |
| ComfyUI official documentation description | Olivio Sarikas | -Zho-blog | |
| ComfyUI_examples | Scott Detweiler |
I share the most practical workflows for free and without reservation to anyone who needs them:
To see the latest ranking of custom nodes, please see this website?nodecafe, built by comfyui-workspace-manager, ranked according to the stars obtained by open source custom nodes !
After testing it myself, I feel that it is a must-install and currently used custom nodes . The ranking is what I think, the number (real-time) is just a reference.
| Custom Nodes (real time) | Introduction (most useful features) |
|---|---|
| ComfyUI | ComfyUI itself is a god-like existence! ComfyUI shortcut keys |
| ComfyUI-Manager | Install, delete, disable and enable various custom nodes of ComfyUI and other functions. |
| ComfyUI-Custom-Scripts | Image source: Image Feed Add a panel that displays images generated in the current session and save your own workflow. |
| comfyui_controlnet_aux | Almost all controlnet preprocessors are integrated into the AIO Aux Preprocessor node |
| rgthree-comfy | Make your workflow clearer, simpler and faster |
| ComfyUI_LayerStyle | Migrate some basic Photoshop functions to ComfyUI, as well as magical functions such as one-click background removal and AI cutout. |
| comfyui-inpaint-nodes | Achieve better (inpainting) effects, including SDXL's Fooocus repair model, LaMa, MAT, etc. |
| ComfyUI-N-Sidebar | The sidebar supports functions such as drag-and-drop, collection, search, and expand/collapse of categories, which greatly enhances the interface interaction experience. |
| ComfyUI-KJNodes | It’s worth it if it has set/get functions. The "noodles" are hidden and the interface is much cleaner. |
| cg-use-everywhere | Wirelessly connect to any interface of the same type, eliminating repeated connections, very practical and refreshing |
| ComfyUI-AnimateDiff-Evolved | Other essentials for animation workflows: - ComfyUI_FizzNodes - ComfyUI-Advanced-ControlNet - ComfyUI-VideoHelperSuite |
| ComfyUI-IF_AI_tools | Enhance your image generation workflow with Ollama using native large language models (LLM) to generate hints. |
| ComfyUI-Easy-Use | The node integration package that simplifies complexity is extended on the basis of tinyterraNodes, and has been integrated and optimized for many mainstream node packages. |
| sdxl_prompt_styler | Style prompt words based on predefined templates. |
| ComfyUI_Fill-Nodes | Fetch images sequentially (on each render) or randomly (also on each render); |
| comfyui_LLM_party | As the front end, comfyui develops a complete node library for LLM workflow construction. It allows users to build their own LLM workflow more conveniently and quickly, and connect it to their own image workflow more conveniently. |
| ComfyUI-enricos-nodes | 8 layers can be dragged, rotated, zoomed, etc. to compose the composition. |
| ComfyUI-Inpaint-CropAndStitch | Automatically crop the mask area → enlarge to the optimal size → redraw, expand and paste back |
If ComfyUI has too many components installed, problems such as environment conflicts may occur. I recommend using Anaconda to create a virtual environment to run ComfyUI to avoid potential conflicts. Even if your virtual environment is installed too much and causes conflicts, you can also create another virtual environment to specifically run the workflow that you need to run now. Specifically, I wrote a detailed document, please criticize and correct me: ComfyUI custom node installation guide in Windows environment, video explanation: The ultimate solution to ComfyUI error!
| Project name and website | introduce | huggingface demo (generated online) | ComfyUI implementation/video introduction |
|---|---|---|---|
| InstantID | [Subject Features] It is possible to use a single picture for ID retention generation | InstantID demo | ComfyUI InstantID: ComfyUI_InstantID-(native support version) |
| SDXL-EcomID | [Main Characteristics] Is Alibaba open source? | SDXL-EcomID | SDXL_EcomID_ComfyUI |
| PuLID | [Main Characteristics] Advantages: high speed + high fidelity + strong style transfer | PuLID demo | ComfyUI-PuLID-ZHO PuLID_ComfyUIcubiq Master Edition |
| IP-Adapter | 【Main Characteristics】 | IP-Adapter demo | ComfyUI_IPAdapter_plus |
| InstantStyle | [Main Characteristics] Strength is style reference | InstantStyle demo | |
| PhotoMaker | Customize realistic photos of people |
| 1. xinsir-sdxl | 2. lllyasviel collection | 3. MistoLine | 4. InstantX-SD3-Controlnet |
|---|---|---|---|
| 5.Shakker-Labs | 6.InstantX | 7. alimama | 8.TheMistoAI |
| 9.jasperai | 10. Apple’s open source depth ml-depth-pro comfyui implementation: ComfyUI-Depth-Pro | 11. Video depth map DepthCrafter | 12.Lotus’ high-quality dense prediction visual basic model based on diffusion, ComfyUI-Lotus |
| Project name and website | introduce | ComfyUI implementation/video introduction |
|---|---|---|
| mochi1 | mochi-1-preview model | ComfyUI-MochiWrapper ComfyUI-MochiEdit |
| LivePortrait | 【Animation】Efficient portrait animation - combined with stitching and redirection control | ComfyUI-LivePortraitKJ |
| ToonCrafter | [Animation] Generate cartoon animation difference, which looks very smooth, ToonCrafter-for-windows | ComfyUI-ToonCrafter |
| GIMM-VFI | The GIMM-VFI model enables generalized continuous motion modeling and the ability to interpolate two adjacent video frames at any time step. | ComfyUI-GIMM-VFI |
| hallo | [Digital Human] Audio-driven portrait, Fudan University, Baidu, Nanjing University | System requirements: Ubuntu 20.04/Ubuntu 22.04, Cuda 12.1? ! |
| MimicMotion | [Animation] High-quality human motion video generation based on confidence-aware posture guidance | ComfyUI-MimicMotionWrapper |
| champ | [Animation, Video] Controllable and consistent portrait animation guided by 3D parameters, this project used Blender! | ComfyUI-champWrapper |
| EchoMimic | [Sound-driven portrait] Alibaba open source | ComfyUI_EchoMimic |
| Omost | [LLM empowers SD] Convert LLM's coding capabilities into image generation (or more accurately, image combination) capabilities, Omost demo | ComfyUI_omost |
|---|---|---|
| ELLA | [Semantic Understanding] Let the large language model LLM empower AI painting | ComfyUI_ELLA |
| DenseDiffusion | [Semantic understanding, deconstructing pictures] Processing dense prompt words, plug and play | ComfyUI_densediffusion |
| florence-2 | [Detection, Segmentation, Description] Microsoft Free | ComfyUI-Florence2 |
| segment-anything-2 | [Detection, Segmentation] Produced by Meta | ComfyUI-segment-anything-2 |
| BrushNet | [Partial redrawing] Partial redrawing has good effect | ComfyUI-BrushNet |
|---|---|---|
| PowerPaint | [Expand, erase, redraw] | ComfyUI-BrushNet with PowerPaint v2 |
| IOPaint | [Erase, redraw, remove background, generate text] A comprehensive, highly completed project | iopaint official website |
| InSPyReNet | Excellent background removal effect | ComfyUI-Inspyrenet-Rembg |
| BiRefNet | Updated with new model in August 2024, BiRefNetV2 | |
| RMBG v2.0 | Updated with new model briaai/RMBG-2.0 in November 2024 |
| flux | 23.8GB full version FLUX.1-dev model download 17.2GBfp8 version FLUX.1-DEV Introduction to Flux.1 | ComfyUI native support |
|---|---|---|
| Kolors | [New model architecture] Kuaishou open source | ComfyUI-KwaiKolorsWrapper |
| PixArt-sigma | [New Model Architecture] A project jointly developed by Huawei's Noah's Ark Laboratory, Dalian University of Technology, and the University of Hong Kong. Personal test available simplified installation tutorial | ComfyUI_ExtraModels Official - How to use PixArt in ComfyUI Workflow 12-PixArt-sigma |
| HunyuanDiT | [New Model Architecture] Tencent Open Source | comfyui-hydit official ComfyUI native support |
| AuraFlow | AuraFlow-v0.2 | ComfyUI native support |
| Stable Diffusion 3 | Apply for API: Stability API key | Official version: ComfyUI-SAI_API |
| CogView3 | Produced by Zhipu |
| SUPIR | [Picture Magnification] The ultimate and excellent magnification solution, blurred photos restore photo-level texture | ComfyUI-SUPIR |
|---|---|---|
APISR | [Image enlargement] - Unofficial implementation of APISR. Super-resolution model specially used for animation, APISR demo | ComfyUI-APISR |
| aura-sr | [Image Enlargement] This is a variant of the GigaGAN paper on amplification under image conditions, used to achieve super-resolution of real-world images. | ComfyUI-AuraSR-ZHO AuraSR-ComfyUI |
| InstantIR | [Image Enlargement and Repair] A new single-image repair model designed to restore damaged images, capable of generating ultimate quality and realistic details. You can also improve the performance of InstantIR with additional text prompts and even achieve customized editing! | ComfyUI_InstantIR_Wrapper |
| res-adapter | [Resolution Adaptive] Plug and play, any diffusion model can generate images without resolution restrictions | ComfyUI-ResAdapter |
| HiDiffusion | [High Resolution] Unlock high resolution creativity and efficiency in pre-trained diffusion models | ComfyUI-HiDiffusion |
| In-Context-LoRA | [Consistency] The core concept of IC-LoRA is to stitch the condition image and the target image into a composite image, while using natural language to define the task. This approach allows it to be flexibly adapted to various application scenarios. | Just how to use ordinary LoRA |
| AnyText | 【Generate Text】Multi-language (including Chinese characters!) visual text generation and editing | AnyText Installation Guide AnyText Video Guide ComfyUI-AnyText |
|---|---|---|
| Glyph-ByT5 | [Generate text] Custom text encoder for accurate visual text rendering | |
| IDM-VTON | [Change] What is the best effect of changing your clothes so far? | ComfyUI-IDM-VTON |
| CatVTON | [Replacement] Relatively lightweight | Comfyui-CatVTON |
| Hyper-SD | [Generation Acceleration] One of the latest State-of-the-Art diffusion model acceleration technologies. | Hyper-SD installation method and workflow Hyper-SD video actual measurement |
| LayerDiffuse | [Generate images with transparency information] Transparent image layer diffusion using potential transparency | ComfyUI-layerdiffuse |
| BasicPBC | 【Comics coloring】 | |
| HivisionIDPhotos | Create ID photos of various sizes with one click, highly integrated, you can experience huggingface online - HivisionIDPhotos | HivisionIDPhotos-ComfyUI |
| InvokeAI | [AI painting complete solution] has a canvas function, which can be generated while painting, which is very suitable for professional people. | invoke official model library invoke official website |
|---|---|---|
| Fooocus | Preview version mashb1t-Fooocus | |
| pinokio | Install, run, and control various AI projects with one click | |
| SwarmUI | Modular Stable Diffusion web user interface |
| facefusion | The author considers himself to be the “next generation face-changing tool” |
|---|---|
| RVC voice changer | RVC, the voice changer I use for videos, (#^.^#). Download the Baidu network disk with a thoughtful address, unzip it and use it! |
| ChatTTS | Text to language, focusing on personification, ChatTTS-uiwebUI version is very good ChatTTS-Enhanced another project. |
| VideoLingo | Translated and cultivated, the author has made some very professional optimizations and is actually usable. |
| pyvideotrans | Translation and dubbing are also being continuously optimized, so keep track of them. |
| marker | [PDF to Markdown] is easy to install and the effect is amazing! Use it to translate and share papers! |
| GOT-OCR2.0 GOT-OCR-2-GUI | All artificial optical signals (such as ordinary text, mathematical/molecular formulas, tables, charts, musical scores, and even geometric figures) are collectively called "characters", and a general OCR theory and an excellent model, GOT, are proposed. , to promote the arrival of OCR-2.0. |
| Developer | Introduction (Main Contributions) |
|---|---|
| ZHO-ZHO-ZHO | BRIA_RMBG 1.4 in ComfyUI: The best background removal model currently developed by BRIA, which already supports batch processing (can remove video background) ComfyUI InstantID: High-quality character retention with just one image! A variety of styles to change as you wish. ComfyUI-BiRefNet-ZHO background subtraction artifact. ComfyUI-Phi-3-mini leverages local large language models. |
| kijai | ComfyUI-SUPIR ComfyUI-BrushNet-Wrapper ComfyUI-IC-Light |
| huchenlei | ComfyUI-layerdiffuse ComfyUI_omost |
| lllyasviel | Fooocus stable-diffusion-webui-forge ControlNet IC-Light |
| ltdrdata | ComfyUI-Manager ComfyUI-Impact-Pack |
| HumanAIGC Ali | Alibaba’s open source projects EMO: Emotional portraits come alive AnimateAnyone: With this, anyone can skip subject three OutfitAnyone: Dress up, online demonstration: OutfitAnyone demo |
| TencentARC Tencent | PhotoMaker T2I-Adapter BrushNet partial redrawing is great! |
| PKU-YUAN-Lab (Yuan Li’s research group-Peking University Information Technology) | Open-Sora-Plan Peking University is still awesome. MagicTime comfyui: ComfyUI-MagicTimeWrapper |
| URL | illustrate | |
|---|---|---|
| OpenArt | [Workflow Sharing] Many big names have shared a large number of high-quality workflows. | |
| civitai | [Models, Workflows, etc.] Search, download and share various AI painting models. Is there any website with the most complete features? | |
| Nerdy Rodent | [Youtube Boss] Teach you step by step how to install the latest AI project using the command line | |
| Latent Vision | [Youtube Boss] Author of ComfyUI_IPAdapter_plus | |
| Olivio Sarikas | [Youtube Boss] Both SD and ComfyUI can have 200,000 fans! | |
| Scott Detweiler | [Youtube boss] The chief quality assurance engineer of Stability.ai. The internal staff is really good. His videos are very good, but all subsequent videos will be charged. |
| ideogram | The images generated by this website are of good quality and can generate text. Their models are also constantly iterating. |
|---|---|
| haiper | [Generating video] Personally, I think the generated effect is better than runway or pika! It's just that 2 seconds is too short. |
| suno audio elevenlabs | [Generate music and sound] chatgpt in the world of music generation? |
| heygen homepage hedra | [heygen] is a cloud video creation platform that provides functions such as video editing ( video translation ), AI-generated content ( digital human ), personalized customization, template scene library, etc., and is suitable for users of various backgrounds and skill levels. 【hedra】Audio drives digital people. |
| Tripo3D Rodin | [Generate 3D model] The 3D model generated is very good and is being iterated continuously. 【3D Generation】Yingmoo Technology, reaches commercial level? |
| Skill | Video commentary |
|---|---|
| 1️⃣ How to share model files in ComfyUI | Lazy people must learn! ComfyUI model sharing settings save space and worry! Lazy people must learn! ComfyUI model sharing settings, saving space and worry!_Ultimate version, the Link Shell Extension download address mentioned in the video |
| 2️⃣ ComfyUI custom node installation guide in Windows environment | The ultimate solution to ComfyUI error reporting! |
| 3️⃣Achieve freedom of translation | Achieve freedom of translation |
| 4️⃣ huggingface download accelerator-webUI | huggingface download accelerator-webUI? Double-click run.bat, then hold down ctrl and click on the URL. If there is a problem with the download, you can switch the proxy off. If you download repeatedly, you need to go to C:Usersc6023.cachehuggingfacehub to delete the cache of the previous download, and then download again. ?(Everyone’s folder is different) |
| shortcut key | illustrate |
|---|---|
| Ctrl + Enter | Add the current graph to the build queue |
| Ctrl + Shift + Enter | Add the current graph to the first position in the build queue |
| Ctrl + Z/Ctrl + Y | Undo/Redo |
| Ctrl + S | Save workflow |
| Ctrl+O | Load workflow |
| Ctrl+A | Select all nodes |
| Alt+C | Collapse/expand selected nodes |
| Ctrl+M | Mute/unmute selected nodes |
| Ctrl+B | Bypass the selected node (equivalent to removing the node from the graph and reconnecting the line) |
| Delete/Backspace | Delete selected node |
| Ctrl + Delete/Backspace | Delete current picture |
| Space | Hold down the space bar and move the mouse to move the canvas |
| Ctrl/Shift + Click | Add clicked node to selection |
| Ctrl + C/Ctrl + V | Copy and paste selected nodes (no connections to unselected node outputs are maintained) |
| Ctrl + C/Ctrl + Shift + V | Copy and paste selected nodes (keeping unselected node outputs connected to pasted node inputs) |
| Shift + Drag | Move multiple selected nodes simultaneously |
| Ctrl + D | Load default image |
| Q | Toggle queue visibility |
| H | Toggle history visibility |
| R | Refresh graph |
| Double-Click LMB | Open the node quick search panel |
| For macOS users, Ctrl can also be replaced with Cmd |
