ComfyUI Custom Extensions

Run ByteDance Lance-3B (unified image/video generation, editing, and understanding) in ComfyUI, with dynamic VRAM streaming for low-VRAM GPUs.

个人零散节点收录，开放给大家用

A package of useful ComfyUI nodes optimizing user experience.

Native ComfyUI nodes for Zen Image Edit with a single-file Qwen-Image-2.1 checkpoint and student text encoder.

A metadata-aware, verified model asset installer for ComfyUI with Xet-accelerated Hugging Face downloads and Civitai analysis.

Provide two custom nodes to load and save images with captions as separate files.

Comfy UI nodes for IMtalker to run native weights. With Automatic model downloads.

A custom node for ComfyUI that provides advanced 2D relighting capabilities.

Nodes: image_layering, color_correction, model_router

Standalone NovelAI image-generation plugin for ComfyUI (V4.5 and later): native NAI prompt scheduler with multi-character prompting, txt2img / img2img / inpainting, Vibe and Precise Reference, metadata reuse, Anlas usage monitoring, and original PNG or lossless WebP saving. Optional Wildcard integration with 4A Prompt Manager.

This repo contains nodes for ComfyUI that implement some helpful operations on tensors, such as normalization.

ComfyUI Memory Management and VRAM Optimization Custom Node - Free VRAM with sequential model loading support

Nodes:ApiRequestNode, LoadVideoNode, JsonParserNode, OllamaRequestNode, OldPhotoColorizationNode, waveform_2_audio, SaveImageUnifiedOutput, VideoHelperUnifiedOutput, ...

Custom nodes to enable ViewComfy app functionalities

A node pack for pausing execution asynchronously and notifying the user

This repo cast Recursive Generalization Transformer for Image Super-Resolution to ComfyUI, the original [a/paper link](https://arxiv.org/abs/2303.06373) and [a/github link](https://github.com/zhengchen1999/RGT)

A matting toolkit based on ComfyUI, supporting multiple matting models and detail processing methods.

A ComfyUI custom node implementing AutoGuidance from the paper Guiding a Diffusion Model with a Bad Version of Itself.

Frequency-aware RoPE reference-attention modulation for Flux/Flux.2 models in ComfyUI.

Frequency-aware RoPE reference-attention modulation for Flux/Flux.2 and MiniMax H3 in ComfyUI.

A collection of custom nodes for ComfyUI that provide dynamic input selection and intelligent upscaling functionality.

ComfyUI nodes for OrpheusTTS model.

A collection of utility / quality-of-life nodes for ComfyUI. Probably only useful to me.

Nodes:visual_anagrams_sample, visual_anagrams_animate

Professional panorama generation tools for ComfyUI converting perspective images/video to equirectangular 360° panoramas with automatic camera calibration via AI, horizon detection, or vertical line detection.

Save and reload ComfyUI latent tensors from the output folder.

ComfyUI模型加载带预览图功能节点 是一款专为 ComfyUI 用户打造的高效可视化模型加载插件，通过直观的缩略图预览和强大的分类搜索功能，将繁琐的模型选择过程转化为愉悦的视觉体验。

Nodes:ImageOverlap-badger, FloatToInt-badger, IntToString-badger, FloatToString-badger, ImageNormalization-badger, ImageScaleToSide-badger, NovelToFizz-badger.

ComfyUI-TrashNodes-DownloadHuggingface is a ComfyUI node designed to facilitate the download of models you have just trained and uploaded to Hugging Face. This node is particularly useful for users who employ Google Colab for training and need to quickly download their models for deployment.

In-browser file manager for ComfyUI — browse, preview, upload, rename, move, copy, and trash files in configured roots

CURRENTLY IN ACTIVE DEVELOPMENT (BETA)! IF THINGS BREAK ITS BECAUSE I BROKE IT.... This my implemenation of a `pipe` in ComfyUI. Is it better or worse than others? No idea. (also has some utility nodes)

These nodes are mainly used to translate prompt words from other languages into English. PromptTranslateToText implements prompt word translation based on Helsinki NLP translation model.It doesn't require internet connection。

A powerful ComfyUI custom node that extracts workflow metadata from PNG images and video files generated by ComfyUI.

Interactive tag attention explorer (SDXL only) Click a spot in the image → the related tag Click a tag → the related part of the image

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.

Quality-focused Triton W8A8 implicit-GEMM Conv3D acceleration for the Qwen-Image/Wan VAE in ComfyUI, optimized for faster first workflow runs and resolution changes. PuppetVisionAI: https://www.youtube.com/@PuppetVisionAI | LinkedIn: https://www.linkedin.com/in/allen-b-3a35505a/ | Website: https://puppetvision.nl

all art styles

ComfyUI custom sampler node implementing Euler A2 sampling algorithm. (Description by CC)

This node for ComfyUI allows saving images with an optional alpha channel (transparency). It supports saving images in formats like PNG, JPEG, and WebP.

A comprehensive collection of color manipulation and analysis nodes for ComfyUI

A ComfyUI custom plugin that scans workflows, detects required models, resolves missing metadata, and installs missing assets directly into the correct ComfyUI model folders.

Yet another set of LLM nodes for ComfyUI (for local/remote OpenAI-like APIs, multi-modal models supported)

A set of nodes for basic Llama 3.2 Vision support in ComfyUI. Give it an image and query and it will output a text response.

Simple custom node for ComfyUI to artificially delay a workflow at any point.

flux patcher for Fill Flux.Dev lora [a/https://civitai.com/models/981615/fluxfill-inpaint-lora](https://civitai.com/models/981615/fluxfill-inpaint-lora)

ComfyUI node of Conversational Speech Model (CSM).

This is a custom node designed to simplify the use of Hunyuan3D in ComfyUI

Automatic image tagging using WD14 models with batch processing and GPU acceleration for ComfyUI

An implementation of Nari-Labs Dia TTS

Local Qwen loader and prompt refiner for ComfyUI

To use stepfun's library, you need an official api that supports multimodal inputs such as video and pictures [a/https://platform.stepfun.com/request-restriction](https://platform.stepfun.com/request-restriction)

Edit embeddings with a curve. Actually should work on any 1D input tensor. Tested with IPAdapter-Plus.

Stacked multi-LoRA loader for MiniMax H3 across bf16, fp8, int8 and w4a8 weights.

A custom ComfyUI node for seamless image editing using fal.ai's NanoBanana and Seedream (v4) models. Edit images with AI-powered prompts, supporting multi-image batches, custom resolutions, and easy fal.ai API key integration directly in the node interface.

Upload images automatically to image hosting sites.

This ComfyUI node will save images directly to Google Drive by using Google's free API service.

Nodes:Send Http Request. You can use this node to save full size images through the websocket.

modify the original node instruction of image vector, add ‘imagemagick’ which is the key base Python library of ‘wand’ library.

OmniGen Unified Image Generation Model Integration.

ComfyUI custom nodes for posting images, fetching model/image metadata, and interacting with Civitai daily challenges via the Civitai MCP server.

This custom node integrates the HunyuanVideo-Foley model for generating audio from video frames and text prompts in ComfyUI. It's built for use in generating Foley sounds from video and text inputs.

Auto-arrange workflow nodes with 4 layout algorithms, group-aware positioning, alignment tools, and undo

ComfyUI nodes for the BFL Flux 3 video API, with an OpenRouter-backed prompt helper.

A Novel AI / SD-WebUI request node, support nai3/nai4, use NovelAI model in Your Comfyui.

Adds 'Assets+' tool to ComfyUI sidebar, which builds upon 'Media Assets' idea albeit expands it to bring Gallery-like experience to Windows machines

A collection of nodes designed for efficiency and the reduction of screen-clutter. Includes a Global Seed controller with boolean toggles, SDXL All-in-One conditioner, a custom SDXL control module, Wildcard processor, Style Injector, and more. [w/WARNING: Updating this node-pack wil overwrite any changes you've made to the included wildcards and styles. Please backup your folders before updating.]

Use [a/Doctor Diffusion's snake oil nLoRAs](https://civitai.com/models/987843) as well as [a/other negative LoRAs](https://civitai.com/models/186617/doctor-diffusions-negative-xl-lora) easily within ComfyUI.

WIP. Nodes: Fetch e621/danbooru image and/or tags etc from a given URL; Get the Wiki entry for a tag through a button press.

A ComfyUI custom node package for seamless integration with Threads (Meta's social platform). This package allows you to publish posts, manage images, and retrieve post history directly from your ComfyUI workflows.

A ComfyUI custom node based on the Volcano Engine Doubao large model Seedream API, designed for high-quality image generation.

A seamless, node-independent way to automatically send your ComfyUI generations to Eagle, complete with full metadata annotation and tags.

ComfyUI nodes for OmniVoice — multilingual zero-shot TTS with voice cloning, voice design, EPUB loading, and voice mixing.

Meta SAM-Audio sound separation with text, time-span, and visual-mask prompts, ComfyUI model management, chunked processing, and optional Comfy Kitchen attention.

A custom ComfyUI node for audio enhancement using Resemble AI's Resemble Enhance model. Created using OpenHands, Cline, and DeepSeek.

ComfyUI-AutoCropBgTrim is a powerful tool designed to automatically clean up the background of your images. This tool trims unnecessary spaces and pixels, leaving only the main subject of the image. It generates both a mask and an image output, making it easy to focus on the essential elements. Perfect for enhancing your photos and preparing them for professional use.

An example node pack for ComfyUI to demonstrate custom node creation.

ComfyUI custom nodes for Seed AI APIs including video generation, image generation, and chat models

A simple but powerful custom node for ComfyUI that patches LoRA models by adding dummy adaLN_modulation_1 weights. This solves compatibility errors when using LoRAs with newer model architectures that expect these keys to be present in the final_layer.

A premium visual enhancement for ComfyUI that adds a smooth 'ghost' effect when dragging nodes and groups.

ComfyUI custom_node for [a/ByteDance's InfiniteYou](https://github.com/bytedance/InfiniteYou).

Geekatplay Studio tiled generative upscale helpers for ComfyUI

Here are several ComfyUI nodes that traverse specific folders to sequentially deliver images, prompts, video files (split into image sequences), entire video files, and video files first split into keyframe sequences at specific intervals to subsequent nodes, which then upload them to the corresponding large models via an API (OpenAI format).

These custom nodes for ComfyUI provide advanced text translation capabilities using Google Translate. They are designed for seamless integration into the ComfyUI environment, offering users powerful tools for text and JSON file translation tasks.

This is a video generation plugin implementation for ComfyUI based on the Lumina Video model.

A practical and lightweight collection of custom nodes for ComfyUI, providing utility tools and convenient workflow functions.

Unofficial implementation of [a/DDColor](https://github.com/piddnad/DDColor) for ComfyUI

Nodes:commonly_node.

ComfyUI node library for fine-tuning LLMs

A ComfyUI node pack for management of larger prompt amounts and prompt variations.

ComfyUI custom nodes for MammothModa2 text-to-image generation.

Nodes for calling LLMs, enabled by LiteLLM

MusePose Remaster is a remaster version of ComfyUI MusePose node. It supports auto weights download, remove most necessary dependencies, etc.

PyTorch extension for ComfyUI featuring extensive PyTorch wrapper nodes for seamless tensor operations and PyTorch model training.

This is an extension script for Stable Diffusion WebUI, modified based on the original functionality. It now supports fixing FLUX panorama seams. It allows users to independently configure seamless image tiling for both the X and Y axes while also being capable of handling FLUX panorama seam issues.

A comprehensive suite of ComfyUI custom nodes for image processing, prompt verification, comparison, extraction, occlusion masking, batch loading, and advanced saving.

A switcher that allows specify conditions and inputs/outputs individually

This package provides ComfyUI nodes for creating 3-scene storyboards from Ollama text output.

A powerful, sleek, and highly optimized LoRA loader for ComfyUI with modern tag management, automated Civitai metadata fetching, and seamless multi-LoRA stacking.

Automatically detects, downloads, and routes missing workflow models into the correct ComfyUI model directories.

A collection of custom nodes for integrating Google Gemini API with ComfyUI, providing powerful AI capabilities for text generation, image generation, and video analysis. Nodes: Gemini Text API, Gemini Image Editor, Gemini Image Gen Advanced, Gemini Video Captioner.