ComfyUI Custom Extensions
Browse 9,420 ComfyUI extensions and 73,215 custom nodes.
A set of nodes to use Reference UNets
This is a set of sampling nodes specifically designed for fine-tuning Twinflow models.
SplatKit World: one image or one sentence -> a 3D Gaussian-splat world you can walk through. A step-by-step app inside ComfyUI on top of ComfyUI-SplatKit: 360 pano, your own drone paths, WAN fly-throughs, SphereSfM, Spirula Studio training, browser viewer.
An unofficial, streamlined, and highly optimized ComfyUI implementation of TeleStyle for video style transfer. (Description by CC)
Visual studio for composing Ideogram 4 structured JSON box prompts in ComfyUI
A collection of simple nodes for Frequency Separation / Frequency Recombine with RGB and HSV methods
Personalize Anything for Free with Diffusion Transformer,use it in comfyUI with wrapper mode
Two-finger scrolling (vertical and horizontal) to pan the canvas. Two-finger pinch to zoom in and out. Command-scroll up and down to zoom in and out. Fixes [a/comfyanonymous/ComfyUI#2059](https://github.com/comfyanonymous/ComfyUI/issues/2059).
ComfyUI Loop Image is a node package specifically designed for image loop processing. It provides two main processing modes: Batch Image Processing and Single Image Processing, along with supporting image segmentation and merging functions.
Nodes:Openpose Editor Plus
Latent bridge for MiniMax-H3 multi-segment continuation: the previous segment's AV latent is spliced into this segment's conditioning — zero re-encoding.
ComfyUI custom nodes for the Agnes AI API — a free, cloud-based AI generation platform. Generate images, create videos, enhance prompts, and analyze visuals — all without a local GPU. Zero extra Python dependencies.
A clean, efficient ComfyUI custom node for VoxCPMTTS (Text-to-Speech) functionality. This implementation provides high-quality speech generation and voice cloning capabilities using the VoxCPM model.
ComfyUI custom node using the fal.ai API for the IC-Light V2 model
Qwen image 2.1 Enhancement Nodes
Per-block and per-layer-type LoRA filtering for Anima and Flux models in ComfyUI.
ComfyUI widget for defining sun light direction for Flux 2 Klein 9B, designed for use with Sun_direction_Lora.
VS Code-inspired MiniMax H3 prompt editor with strict prompt sections, completions, reference media previews, and plain STRING output.
Native Qwen3-VL integration for ComfyUI with image/video analysis and uncensored model support.uilt-in presets for various styles: *Descriptive, Danbooru/e621 tags, Art Critic, Social Media, NSFW,* Batch & Loop Support,Next Scene Generation,JoyCaption Integration
A custom ComfyUI node for testing multiple LoRA models across different strength values. Automatically generates model/clip pairs for each LoRA-strength combination and includes an image saver node with text overlay for easy identification.
Pixel art Enhancement Node for ComfyUI
This custom node package provides nodes specifically for using the mvadapter_ig2mv_sdxl.safetensors adapter within ComfyUI. This adapter is designed for image-guided multi-view generation, typically used for creating textures from 3D mesh renders (position and normal maps).
ComfyUI Nodes that generate prompts and many more.
Template repository for custom nodes from [ComfyUI-CustomNodes-Template](https://github.com/jhj0517/ComfyUI-CustomNodes-Template).
Multi-reference image conditioning nodes for LTX-2.5 in ComfyUI, using MSR slot embeddings and negative temporal offsets.
Detect human parts using the DeepLabV3+ ResNet50 model from Keras-io. You can extract hair, arms, legs, and other parts with ease and with small memory usage.
Adds a gallery to the Load Image node and tabs for Load Checkpoint/Lora/etc nodes
Run Tencent's HunyuanImage-3.0 (80B MoE) natively in ComfyUI on a 24 GB GPU: base, Instruct and Instruct-Distil, prompt rewriting, image editing, Spectrum speed-up.
TQ3 KV cache compression for ComfyUI reducing attention KV cache VRAM by ~4.5x using 3-bit Lloyd-Max quantization with Fast Walsh-Hadamard Transform decorrelation.
Checkpoint and UNET/Diffusion Model loaders for <a href="https://github.com/bitsandbytes-foundation/bitsandbytes">BitsAndBytes</a> NF4 quantized models
Nodes for loading both Checkpoints and UNET/Diffussion models quantized to bitsandbytes NF4 or FP4 format. Still under development and some limitations such as using LoRA might apply still.
StableAvatar: Infinite-Length Audio-Driven Avatar Video Generation,you can try it in ComfyUI
Multimodal prompt studio for ComfyUI. Write and refine prompts for image, video and audio models using local or API-based LLMs.
VantageLongWanVideo is a lightweight project management system for ComfyUI, designed for long-running, prompt-driven video generation. It simplifies workflow management by saving prompts to disk, handling continuity across runs, and allowing resuming or regenerating from any point.
A workflow-compatible NVIDIA RTX Video Super Resolution Pro node for ComfyUI with exact 16K hybrid processing and channel-corruption detection.
千问 2.1(Qwen-Image 2.1)优化节点包:优化加载器(精度档/turbo LoRA/采样位移/前缀缓存)+ 加速链(BlockCache T8 / TE-Speed / Speedup / RDNA3-FA,按互斥规则自动挑)+ 分块解码(防大图 OOM)+ ControlNet + 一键出图。缺文件/缺节点会如实跳过并给出说明。
A ComfyUI integration for FireRedTTS‑2, a real-time multi-speaker TTS system enabling high-quality, emotionally expressive dialogue and monologue synthesis. Leveraging a streaming architecture and context-aware prosody modeling, it supports natural speaker turns and stable long-form generation, ideal for interactive chat and podcast applications.
Semantic conditioning bridge for MiniMax H3.
A workflow subscription and version management hub for ComfyUI.
ComfyUI nodes for the different projection models and camera movements
Intel Arc-first ComfyUI monitor with real-time GPU/CPU/RAM stats and an exclusive workflow execution success rate predictor. NVIDIA (CUDA) fully supported.
Dashboard Nodes for Comfy
General-Purpose Multi-Backend PyTorch LoRA Trainer for ComfyUI (Intel Arc XPU, ROCm, CUDA, Cloud TPU, Ascend NPU, Gaudi HPU, DirectML, MPS, CPU).
ComfyUI-WorkflowGenerator custom nodes for generating ComfyUI workflows from natural language
Heuristic modification of the Heun sampler using a custom function based on normalized distances. For ComfyUI.
Provides the ability to set the temperature for both UNET and CLIP. For ComfyUI.
Wrapper nodes for OpenDiT: [a/OpenDiT](https://github.com/NUS-HPC-AI-Lab/OpenDiT/), supports Open-Sora t2i and i2i
A Jupyter-style custom node for executing Python code within ComfyUI workflows. Features syntax highlighting, shared variables between cells, and pre-loaded libraries (numpy, torch, PIL, matplotlib).
Nodes to use [a/Smoothed Energy Guidance](https://github.com/SusungHong/SEG-SDXL) for ComfyUI.
ComfyUI-TinyBreaker is a collection of custom nodes specifically designed to generate images using the TinyBreaker model. It's actively developed with ongoing improvements. Although still in progress, these nodes are functional and allow you to explore the potential of the model.
A single-purpose, multi-channel curve editor for ComfyUI, providing precise color control over R, G, B, and Luma channels directly within the node graph. It’s a focused, lightweight, and standalone solution built specifically for one task: applying color curves cleanly and efficiently.
ComfyUI custom node implementing Florence 2 + Segment Anything Model 2, based on [a/SkalskiP's HuggingFace space](https://huggingface.co/spaces/SkalskiP/florence-sam)
ComfyUI custom nodes for LongCat Video Avatar 1.5 audio-driven human video generation.
Transparent video from MiniMax H3 in ComfyUI with keying studio and ProRes exports.
ComfyUI内で見本から漫画のコマ割りを選び、人物・場面・セリフをMOSSへ渡すカスタムノード
A custom ComfyUI node designed to create seamless motion effects from single images by integrating with Hunyuan Video through latent space manipulation.
Crosshair guidelines to help align nodes and groups while moving or resizing.
ComfyUI nodes for video object segmentation using [a/SAMURAI](https://github.com/yangchris11/samurai) model.
Nodes: DeepFloyd, Filter, Select, Save, Decode, Encode, Repeat, Noise, Noise
Implementation of color transfer using KMeans algorithm
一个 ComfyUI 扩展,通过 API 集成实现图像生成及相关功能
This extension contains a custom node for ComfyUI. The node, called 'Bounding Box Crop', is designed to compute the top-left coordinates of a cropped bounding box based on input coordinates and dimensions of the final cropped image. It does so computing the center of the cropping area and then computing where the top-left coordinates would be.
TK Toolkit (formerly Anima Toolkit) — all-in-one Anima/SD toolkit for ComfyUI: LoRA management (batch apply + Civitai metadata/downloads), Danbooru search & waterfall gallery, bilingual prompt cards with Chinese tag autocomplete, an output gallery over ComfyUI/output, and batch workflow utilities.
Structured photorealistic prompt builder for FLUX Klein 9B in ComfyUI
A custom node for ComfyUI. It compare two images to rate facial similarity.
The official Draw Things extension for ComfyUI. Sends image-generation requests to Draw Things over gRPC. Supports Bridge Mode for DT+ cloud and local model execution.
An Unofficial ComfyUI custom node package that integrates [a/Diffusion Self-Distillation (DSD)](https://github.com/primecai/diffusion-self-distillation) for zero-shot customized image generation. DSD is a model for subject-preserving image generation that allows you to create images of a specific subject in novel contexts without per-instance tuning.
experimental node pack to test using reference videos for their motion.
Intelligent path organization for ComfyUI. Drag-drop segments, auto-detection (Model, LoRA, Resolution, Seed), template variables, global presets, dual outputs for Image Saver, and 7 beautiful themes. Say goodbye to messy output folders!
Quality-of-life tools for ComfyUI, built on the V3 API with full subgraph and Nodes 2.0 support: a Flow Visualizer that shows how data flows through a workflow and replays the last run (what ran, was cached, skipped, failed, or changed since), a resource monitor (CPU/RAM/GPU/VRAM/temp/power with history graphs), per-node timing badges, node search with jump-to-node and Follow Execution, graph auto-arrange (Quick/dagre/ELK), and handy nodes: Profiler Timing, Load Image (With Subfolders), Parse JSON, Execution Gate, Play Sound and System Notification. A modern alternative to ComfyUI-Custom-Scripts and Crystools.
The 'Eses Image Effect Levels' is a ComfyUI custom node that provides a real-time levels adjustment tool directly within the user interface. It allows for interactive control over the tonal range of both images and masks, using a GPU-accelerated PyTorch backend for near instant feedback.
A semi-random prompt generator for danbooru tags that works alongside your character prompt, allowing you to put your waifu in many scenarios. (for Illustrious/NoobAI/Pony)
ComfyUI implemntation for [a/SyncTalk](https://github.com/ZiqiaoPeng/SyncTalk)
This repository contains ComfyUI custom nodes for StepVideo.
Provide ComfyUI support for FramePack start-and-end image reference
Nodes:Conditioning (Blend), Inpainting VAE Encode (WAS), VividSharpen. Experimental nodes, or other random extra helper nodes.
Lightweight alignment toolbar for ComfyUI graphs. Provides quick align, equal size, and distribution actions with optional keyboard shortcuts.
This repository provides integration of GPT-4 and Claude 3 models into ComfyUI, allowing for both image and text-based interactions within the ComfyUI workflow.
Attempts to implement [a/CADS](https://arxiv.org/abs/2310.17347) for ComfyUI. Credit also to the [a/A1111 implementation](https://github.com/v0xie/sd-webui-cads/tree/main) that I used as a reference.
Dynamic Node examples for ComfyUI
Custom ComfyUI node for Qwen3.5-9B — a unified natively multimodal model with image, video, and text understanding capabilities including thinking mode.
MiniMax Reference Director — ComfyUI custom nodes for subject management, timeline editing, and VLM-enhanced prompt generation.
Visual image editing, Anima regional prompting, and MOSS-Image layout prompting for classic ComfyUI.
Integration framework for KIE Nano Banana Pro API into ComfyUI custom nodes with API key setup and placeholder for upcoming node implementations. (Description by CC)
Nodes: MS kosmos-2 Interrogator, Save Image w/o Metadata, Image Scale Bounding Box. An implementation of Microsoft [a/kosmos-2](https://huggingface.co/microsoft/kosmos-2-patch14-224) image to text transformer.
ComfyUI custom nodes for SoulX-FlashHead generating real-time streaming talking head videos from reference images and audio input.
ComfyUI adaptation of https://github.com/G-U-N/Motion-I2V
tag completer with csv file
Save a picture as Webp file in Comfy + Workflow loading
HDM model loader for ComfyUI
Reference images for Krea 2 via native Qwen3-VL vision tokens: character and style transfer with the companion reference LoRA (layered loader included).
A ComfyUI custom node for automatic video scene segmentation using TransNetV2.
Nodes: Runtime44Upscaler, Runtime44ColorMatch, Runtime44DynamicKSampler, Runtime44ImageOverlay, Runtime44ImageResizer, Runtime44ImageToNoise, Runtime44MaskSampler, Runtime44TiledMaskSampler, Runtime44IterativeUpscaleFactor, Runtime44ImageEnhance, Runtime44FilmGrain
ou can using DeepFakeDefenders in comfyUI to Prediction image is a DeepFake img or not.
SparkVSR: Interactive Video Super-Resolution via Sparse Keyframe Propagation
Spectrum: training-free diffusion sampling acceleration via Chebyshev polynomial feature forecasting. Drop-in KSampler replacement that skips transformer blocks on predicted steps for ~2-3x speedup.
ComfyUI-Qwen-Omni is the first ComfyUI plugin that supports end-to-end multimodal interaction, enabling seamless joint generation and editing of text, images, and audio. Without intermediate steps, with just one operation, the model can simultaneously understand and process multiple input modalities, generating coherent text descriptions and voice outputs, providing an unprecedentedly smooth experience for AI creation.
把 ComfyUI 的生成流程放进一张无限画布:节点即素材,拖线即派发,支持真实进度、资产库与提示词库
This is a hierarchical auxiliary project of the IPAdapter project, which uses a slider to quickly control the hierarchical weights and add fully random and semi random modes