ComfyUI Custom Extensions
Browse 9,420 ComfyUI extensions and 73,215 custom nodes.
Nodes to level up your workflows performance and streamline specific functions.
A comfyui costume node by BillBum for using api gen (VLM LLM T2I API Tools)
ComfyUI latent-space post-processing for recurring GPT Image texture artifacts
A collection of custom nodes designed for ComfyUI from the AJO-reading organization. This repository currently includes the Audio Collect & Concat node, which collects multiple audio segments and concatenates them into a single audio stream.
Advanced live tuning nodes for SDXL, Illustrious, and NAI. Slice U-Net and CLIP into functional areas to adjust intensity of specific components like Layout, Style, or Syntax.
A ComfyUI custom node that enhances image and video generation prompts using LM Studio's local OpenAI-compatible API.
Capture Webcamera and URL media streams as ComfyUI images.
ComfyUI Nodes for using Spout streams.
Ultralytics workflow and editor tools for ComfyUI with enhanced preview and UI integration.
Faithful volumetric video of real performances — surround coverage from multi-camera capture, engineered for measurable stability.
A ComfyUI custom node based on OpenRouter that supports calling the Google Gemini model via the OpenAI API to generate new images from reference images and prompts.
A small collection of custom nodes for use with ComfyUI, for geometry calculations
ComyUI wrapper for Intel OIDN image denoising
Nodes: LoRA-Tuner. For using multiple LoRA easily.
Automatically creates checkpoints during workflow execution. If If an workflow is canceled or ComfyUI crashes mid-execution, then these checkpoints are used when the workflow is re-queued to resume execution with minimal progress loss.
Keep faces when you regenerate MiniMax H3 videos to 2K with NVIDIA's Open-H3-Regenerate-2K: an automatic face mask (OpenCV YuNet) makes faces start the refine from lower noise while the rest of the frame gets the full regenerate.
Dia TTS model capable of generating ultra-realistic dialogue in one pass. ComfyUI node.
Experimental sampler node. Sampling alternates between A and B inputs until only one remains, starting with A. B steps run over a 2x2 grid, where 3/4's of the grid are copies of the original input latent. When the optional mask is used, the region outside the defined roi is copied from the original latent at the end of every step.
HeartMuLa audio generation and lyric transcription
Google AI Studio by BuffMcBigHuge
Nodes: XY Plot with many options, Output Config (JSON / JSON5 / YAML), Queue, Format String, List utilities, Input nodes, ....
Bytebeat is like composing music with the tools of a programmer’s toolkit. Instead of piano keys, you have operators like >>, |, and &. It’s like giving your CPU a guitar and letting it shred! 🤘
Rectified-flow samplers for Z-Image-Turbo. Clean interpolation formula without sigma division, linear schedulers, higher-order variants. Fixes built-in euler/euler_ancestral instability issues
Optimize prompts for text-to-image models at no cost.
Custom sigma scheduler node for ComfyUI. Easily tune, visualize, and concatenate dynamic sigma curves with an interactive graph.
One-node NegPiP prompting for Z-Image, SD1, SDXL, and Anima in ComfyUI.
Provide various custom nodes for Latent, Sampling, Model, Loader, Image, Text. This is the fixed version of the original [a/ComfyUI-nodes-hnmr](https://github.com/hnmr293/ComfyUI-nodes-hnmr) by hnmr293.
Slightly better random prompt generation tools that allow combining and picking prompts from both file and text input sources.
Speaks scripts with your trained GPT-SoVITS character models (v1 to v2ProPlus; Japanese, Chinese, English): emotion tags, several speakers in one script, pauses. No separate GPT-SoVITS install needed; base models download on first use. Companion of Anomalous Model Browser (import characters, edit emotions and pronunciations), also usable on its own.
A set of custom nodes that I've either written myself or adapted from other authors
porting audioreactivity pipeline from vktrs to comfyui.
Nodes:Whisper Node, Prompt Schedule Converter. Convert song lyrics into a useable prompt travel schedule within comfyUI. Includes whisper large-v2.
A powerful workflow scheduling extension for ComfyUI that enables automated daily execution of workflows with an intuitive web interface ,Adding shutdown computer after workflow node
A standalone ComfyUI custom node package for Facebook's MusicGen using Hugging Face Transformers. Generate high-quality music from text prompts with full support for CUDA, MPS (Apple Silicon), and CPU.
lucky_seed
A custom ComfyUI node that implements A1111-style prompt handling with proper isolation and emphasis math.
Timed and untimed Qwen-only image and video references for ComfyUI's core MiniMax H3 Ref2VA node.
NODES:Nsfw Image Check Node
Subject manager, subject crafter, asset manager
fcFloatMatic is a custom module, that when configured correctly will increment through the lines generating you loras at different strengths. The JSON file will load the config.
Original author is ByteDance. ComfyUI sampler for HyperSDXL UNet Ported from: [a/https://huggingface.co/ByteDance/Hyper-SD](https://huggingface.co/ByteDance/Hyper-SD)
A custom node for ComfyUI to create a prompt based on a list of keywords saved in CSV files.
Adds a browse button to every model dropdown so you can load models from anywhere on disk. Extra paths persist across restarts. Supports English, Spanish, Chinese and Japanese.
Tiled upscale and refine for reference-latent edit models (FLUX.2 klein and similar): regenerate detail beyond the resolution a model can handle in one pass, without visible seams.
WorkflowX bundles ComfyUI configuration profile nodes with XFlows workflow management, XPrompts prompt libraries, XNodes reusable node snippets, and the JsonX prompt system.
A modern re-implementation of the classic [a/cg-image-picker](https://github.com/chrisgoringe/cg-image-picker) ComfyUI nodes. Keep the workflow-pausing image selection experience while shedding the legacy compatibility shims.
PixRestore-S one-step image restoration for ComfyUI: 512x512 RGB, official path reproduced bit-exactly (eager mode), pinned and hash-checked upstream files.
This is a Seed-X-PPO-7B ComfyUI plugin. Easy to use
A faster tiled VAE decode node for ComfyUI based on LTXV, optimized for low VRAM scenarios with configurable tiling, progress bars, and improved logging for decoding large images and videos.
Professional ComfyUI custom nodes for managing LoRA trigger words, with auto-read, smart lookup via Civitai, editing, and WebUI-compatible JSON format support. (Description by CC)
ComfyUI custom user.css and some script stuff. mainly for web interface.
GGUF Quantization support for native ComfyUI models with FantasyTalking.
various deforum style effects
A custom node for ComfyUI that analyzes and sorts images based on head orientation using MediaPipe. It detects facial landmarks, calculates head pose, and intelligently sorts images for enhanced AI image processing workflows.
Ports the official ComfyUI GPT-API node, adding support for customizable api_base, auth_token, and model settings.
ComfyUI nodes to use [a/DiLightNet](https://github.com/iamNCJ/DiLightNet). These nodes can run DiLightNet, but the Dust3r or BlenderPy implementations to create lighting are not included. Expect those to be added to seperate repos when time allows.
Token-aware spatial Multi-LoRA composition and transparent supersampling for Krea 2 in ComfyUI.
A Anime Character Segmentation node for comfyui, based on [this hf space](https://huggingface.co/spaces/skytnt/anime-remove-background).
一键把当前 ComfyUI 工作流推到 Modal Serverless GPU(H100)跑,自动同步本地模型/自定义节点,出图回填画板,零终端。 | One-click run the current ComfyUI workflow on Modal Serverless GPU (H100); auto-sync local models/custom-nodes, results back to the canvas, no terminal.
This extension provides a custom node that allows the use of [a/Matte Anything](https://github.com/hustvl/Matte-Anything) in ComfyUI.
ComfyUI-Unwatermark: A ComfyUI custom node to intelligently remove watermarks from images using the unwatermark.ai API. This custom node for ComfyUI allows you to easily remove watermarks from your images by leveraging the power of the unwatermark.ai API.
A node to pause execution of the workflow and select which input images should proceed to the output.
A training-free ComfyUI model patcher that accelerates diffusion sampling via Chebyshev spectral forecasting, compatible with standard samplers like Euler and DPM++. (Description by CC)
Enhance ComfyUI workflow by providing powerful node search, attribute search, and port navigation features with Ctrl+F shortcut support.
This extension contains various nodes for CFG scheduling and more. NODES:ScheduledCFGGuider, PerpNegScheduledCFGGuider, CosineScheduler, GaussianScheduler, LogNormalScheduler, InvertSigmas, ConcatSigmas, OffsetSigmas, SplitSigmasByValue
Sleek, modern resource monitors (CPU / RAM / GPU / VRAM / Temp / Disk) for the ComfyUI toolbar, with live sparklines.
Quantization tools are from [a/https://github.com/casper-hansen/AutoAWQ](https://github.com/casper-hansen/AutoAWQ) and [a/https://github.com/IST-DASLab/marlin](https://github.com/IST-DASLab/marlin) Only applicable to graphics cards with sm_80 and above (30 series and above) Need to install marlin dependencies first
RT HeartMuLa is a powerful suite of multilingual AI music generation nodes for ComfyUI that enables the creation of professional-quality full songs with lyrics directly on your local hardware.
Custom ComfyUI nodes using CLIP Vision models for image similarity analysis.
All-in-one enhancement plugin with Checkpoint & LoRA Stack and Prompt Enhancement nodes. Customizable multi-dimension tag library, anime-gen specialized. LoRA groups with inline trigger words — pick tags, 1-click quality output. LLM smart prompt expansion, no more prompt block.一个ComfyUI 全链路增强插件,包含底模&Lora栈与Prompt增强节点,内置可自定义的全维度标签库,二次元出图特化,LoRA 群组管理支持触发词内嵌,挑选标签一键优质出图,LLM 智能扩写摆脱词荒
A single ComfyUI node for executing arbitrary Python code with arbitrary inputs of any types.
One hardware-aware loader for the whole MiniMax H3 Director model stack — model, CLIP, VAEs, LoRAs, and attention patches — auto-selected from your detected VRAM. Routes model loading through an adaptive VRAM reservation system that measures real usage per render shape and tightens itself over repeat runs, rather than a fixed guess.
Provides nodes geared towards using GIMP as a frontend for ComfyUI.
Radial marking menu for ComfyUI. Alt+Click to open, hold-drag-release to select and place nodes.
ComfyUI custom_node for ByteDance's InfiniteYou
Typed decisions (noul / choice / score) with calibrated probabilities, on ComfyUI's text-encoder engine. Supports imajev, d3 and Clef-flash.
TTS with emotional speech capabilities in 8 Languages 24 speakers.
www.nuvulabs.ai | Install missing models & custom nodes, production ready VFX & film subgraphs and workflows for secure offline generation, and dependency installations including SageAttention and Triton.
Concept Steering for ComfyUI — steer image generation with learned concept directions via SAE + DPO
Orion4D Anaglyph is a high-performance custom node designed to transform 2D images into stereoscopic renders via a depth map. It offers total control over parallax, convergence, and depth processing to ensure optimal visual comfort
lollms_nodes_suite is a set of nodes for comfyui that harnesses the power of lollms, a state-of-the-art AI text generation tool, to improve the quality of image generation.
Implementation of paint-by-example on ComfyUI
ComfyUI-LBMWrapper is a user-friendly interface designed to simplify the integration of lightweight models into your projects. It streamlines workflows by providing essential tools for managing and deploying models with ease.
Color-coded and nested groups for open ComfyUI workflow tabs.
Easy Pony is a helper node that simplifies the process of adding scoring and other attributes to prompts when using Pony models.
ComfyUI nodes for MiniMax-H3: low-RAM image generation and editing, plus video and audio tools for canvas planning, prompt checking, latents, audio reroll, and NF4 loading.
Open-vocabulary prompt segmentation for ComfyUI powered by YOLOE-26.
Injects three new linking behaviors to rgthree's Fast Groups Bypasser and Fast Groups Muter nodes.
A custom node for ComfyUI that allows you to load images directly from Google Street View to use as backgrounds, textures, or inputs in your workflows.
Custom TTS node that clones voice from a reference audio and speaks entered text.
Toggle torch.backends.cudnn.enabled, or dynamically wrap VAE decoding nodes with cuDNN disabling flags when on AMD (more fast is more good).
View general image properties of image: Dimension, Resolution, Aspect Ratio, Tensor Size (including batch)
Toolkit for creating embeddings for various models in ComfyUI.
ComfyUI nodes for palette-card image conditioning and structured palette prompts.
DPM++ 2M alt Sampler for ComfyUI (And WebUI) and Align Your Steps Scheduler
VisualCloze: A Universal Image Generation Framework via Visual In-Context Learning,you can use it in ComfyUI
Generate parametric 3D heads by adjusting gender, race, expression, and head angle with sliders, output as color images, normal maps, depth maps, and masks for ControlNet conditioning. (Description by CC)
A ComfyUI custom node for running NVIDIA PersonaPlex - a real-time, full-duplex speech-to-speech conversational AI with persona control.
This is a set of custom nodes for ComfyUI designed for generating sequential long video clips based on the Wan 2.2 model architecture, handling continuity between segments and offering advanced control over the sampling process
A custom ComfyUI extension to compare two video/image sequences via wipe preview.