ComfyUI Custom Extensions
Browse 9,446 ComfyUI extensions and 73,215 custom nodes.
Custom ComfyUI nodes to use Jina-clip IP-adapter for Mugen
Native ComfyUI loader for Alibaba PAI MiniMax-H3 PDD acceleration LoRAs
Natural language prompt editing and workflow enhancement tools for ComfyUI.
ComfyUI-Ideogram4-Luna is a custom node pack for building structured Ideogram 4 JSON prompts with a visual canvas editor supporting bounding boxes, regions, colors, text elements, and import/export.
ComfyUI-StylePromptLibrary is a custom node pack for selecting reusable image style descriptions from a 306-entry JSON library with cascading pickers, live search, and style strength controls.
Visual free crop and horizontal straighten rotate node for ComfyUI.
Hold a hotkey to save and paste node selections as reusable snippets. Quick Paste / ComfyUI-QuickModulePicker.
Crazy Logic & Persistent State for ComfyUI. Simplified Tools for complex workflows
A LoadImage-style node that auto-detects the OS clipboard image, shows a live thumbnail, and outputs it as IMAGE/MASK.
Depth-map paper-cut and parallax layer rasterizer for ComfyUI
ComfyUI custom nodes: luminance masks, smart saturation, torch-compile upscale and misc image tools
Professional VRAM monitoring overlay for ComfyUI that displays real-time GPU memory usage as an independent, non-intrusive window. (Description by CC)
Generate images, video, and audio in ComfyUI with Mage models such as Mango, Cherry, and Seed Audio, through the Mage API.
ComfyUI custom nodes for Kling AI image generation, text-to-video, image-to-video, advanced custom elements, motion control, lip sync, text/video to audio, effects, virtual try-on, and image expansion.
A custom node for ComfyUI designed to load, organize, and preview text prompts from local .txt files with support for multiline prompts and real-time dynamic previews.
Auto face detection + inpainting node using SAM3
A custom node extension that adds project management functionality to ComfyUI.
Save animated GIF format nodes in ComfyUI
A custom node pack for multi-LoRA stacking and dynamic weight management
Display real-time mouse coordinates in graph space. Shows the current mouse position in the top-right corner of the canvas, useful for positioning nodes precisely, debugging layouts, and measuring distances.
Set of custom nodes for ComfyUI that generates images using the Flux 2 Pro model via Replicate API with flexible aspect ratios and secure API key management. (Description by CC)
Modular ComfyUI nodes to run the vision-language model MiniCPM‑V‑4 in GGUF format, powered by llama‑cpp‑python.
A powerful ComfyUI custom node that performs object detection using YOLOv8 and outputs individual images for each detected object. Perfect for automatic mask generation, object isolation, and batch processing workflows.
Custom ComfyUI nodes for generating outfit prompts representing diverse ethnicities, cultures, and regions. Uses extensible JSON data for clothing, accessories, and poses, with “random/disabled” dropdowns for flexibility. Advanced prompt engineering is supported via Ollama LLM integration. Easily add new regions, ethnicities, or cultures by updating data files and creating lightweight node wrappers. Designed for artists, researchers, and developers seeking culturally rich, customizable prompt generation in ComfyUI workflows.
MiniMax-H3 video with sound in one node: prompt, an optional first and last frame, and nothing that can be set wrongly. Core-only, no second pack to install.
ComfyUI custom nodes for CatsAPI (Cat Shadow Studio), providing image and video generation nodes with model-specific parameters and API authentication. (Description by CC)
Nodes for modifying a prompt to create prompt variations. ScramblePrompts [m9]: Reorder prompts, remove prompts, modify weights TweakWeights [m9]: Modify the weights of prompts matching keywords
ComfyUI custom node that patches loaded models (Flux, SDXL, etc.) to run Linear layers in FP8 or MXFP8, unlocking tensor-core throughput on Ada Lovelace (RTX 4090) and Blackwell (RTX 5090) GPUs.
Nucleus Image (MoE diffusion) support for ComfyUI — backport of PR #13471
ComfyUI custom nodes for Mistral's Voxtral TTS
Custom ComfyUI nodes for installing Python packages directly from workflows, including wheel installation and pip package management.
Animated metaball halftones and feature-tracking HUD boxes with labels, connectors, native VIDEO and overlay masks.
Universal zero-VRAM diffusion streaming extension for ComfyUI. Enables running Wan2.1 (14B), FLUX.1 (12B), and HunyuanVideo (13B) on consumer Apple Silicon Macs (MLX) and NVIDIA RTX GPUs via 3-stage offloading and block-by-block NVMe streaming.
プロンプトの並びを、公式推奨の順番に並べ替えてから次のノードに渡します。 Stable prompt tag ordering for ComfyUI with model presets, pinned tags, and user dictionaries.
Custom nodes for LTX-Video support in ComfyUI - registry fix - forked by Mattabyte
A ComfyUI extension for the Insta360 DAP (Depth Any Panorama) model providing high-quality depth estimation optimized for panoramic (equirectangular) images.
Multi-point white balance and exposure correction for e-commerce product images.
ComfyUI nodes for LightOnOCR-2-1B, LightOn's 1B-parameter OCR vision-language model. Extracts clean text from documents, receipts, tables, and more.
Turn a 3D model of a building into social-ready campaign assets: geometry-faithful renders, an automatic quality gate, validated copy and a laid-out post — all in one ComfyUI graph.
One rest-pose mesh to a registered keyframe sheet an auto-rigger can animate from.
Nodes:Loop, LoopStart, LoopEnd, LoopStart_SEGIMAGE, LoopEnd_SEGIMAGE
ComfyUI custom node that protects flat, smooth regions (sky, walls, skin) from upscaler hallucination and colour drift -- an edge-preserving high-frequency mask for upscale and refine passes.
Load an image with its real path (system file dialog), compare the result with the original, save it next to the source (name_1, name_2 …) or send it back to the loader and keep editing.
ComfyUI custom node: upscale and crop in one. Five ways to ask for a size -- percent, factor, shortest side, longest side, megapixels -- all aspect-preserving and all working downwards too, plus a crop stage with the classic formats. The upscale model is picked by a widget, so there is no loader node to wire up.
Blender Actions in Comfy using native BPY
A ComfyUI custom node that outputs the current workflow filename as a string.
MQ util nodes for ComfyUI
ComfyUI node to choose target dimensions based on closest aspect ratio
A node that sequentially executes a list of prompts line-by-line from a single text box.
A collection of useful custom nodes that help with a wide range of tasks.
Nodes:INSPY removebg ModelLoader, INSPY RMBG
ComfyUI node that removes skin highlights (specular shine) from anime/realistic character images using blur-brush style processing with AI face exclusion and lineart protection.
use CLIP to convert OKLCH color into prompt color text
ComfyUI node that frees VRAM and RAM by wiping a single model mid-workflow (no CPU copy), or offloading it to CPU.
a set of nodes to help u run ai code using MLTask
Reproduces Qwen TeleStyle's style-transfer pipeline as native ComfyUI nodes with weights read directly from ComfyUI's model folders.
ComfyUI custom node pack: POS label composition and printing (ESC/POS graphic page) for Custom TPTCM60/112 kiosk printers.
ComfyUI custom node pack: pixel-art and retro display filters (pixelization, palettes, dithering, CRT and analog artifacts).
SDLI (Stable Diffusion Latents in Imagefile) is file format, that contains latents and lossy (decoded) image. Detailed format is written at [a/SDLI Tools](https://github.com/MitoshiroPJ/sdli_tools).
The system selects and loads preset.
A dependency-safe and memory-conscious ComfyUI custom node that runs basic Transformer LLMs directly inside ComfyUI workflows.
An image mask painter/editor node that enables users to draw/paint/edit a mask on an external canvas by inputting an image, outputting a mask for further use. Please turn off ComfyUI and reopen it when first added this node, it will then function as normal.
A dynamic group bypass controller for ComfyUI workflows.
A simple image and video resolution selector for ComfyUI.
Crop an image/mask to its bounding box and resize, with an added `multiple` option to force output dimensions to a chosen step size.
Schema-driven ComfyUI gateway for fal.ai — nine unified nodes covering text/image/reference-to-video, text/image/reference-to-image, upscale, plus T2T and I2T. Auto-populates from fal's model catalog (and OpenRouter's vision-capable LLMs for I2T). New models appear in the dropdown automatically — no per-model code required.
Chandra-OCR-2 is a state-of-the-art OCR model that outputs structured markdown, HTML, or JSON while preserving precise layout information from images and PDFs across 90+ languages.
Hrio Commerce Banana image/video generation nodes for ComfyUI, with normal single-output image node, normal single-output video node, and panel editor.
Prompt diagnostics output node for elapsed time plus sampled peak process RAM and device VRAM.
Small ComfyUI nodes that remove needless workflow friction.
A comprehensive collection of utility and quality-of-life nodes for ComfyUI workflows
A custom node for ComfyUI that implements the Jump Flood Algorithm (JFA).
A ComfyUI custom node for depth-based geometric view synthesis with 6-DOF camera transforms, parallax effects, and hole mask output for inpainting.
A collection of simple utility nodes for ComfyUI
This plugin works with [a/IF_AI_Tools](https://github.com/if-ai/ComfyUI-IF_AI_tools) to build a workflow in ComfyUI that uses AI to assist in generating prompts.
Encode Prompt in TOML for ComfyUI.
Trimmed English fork of Comfyui-Qwen-Image-2.1-MultiAngle-T8: interactive 3D camera studio for Qwen Image 2.1 with TripoSplat reconstruction.
MindMap nodes are used to build knowledge graphs within ComfyUI workflows.
This extension offers a custom node to save image to S3-compatible oss.
Pads images to multiples of N (right/bottom only) with solid color fill. Useful for VAE/model dimension requirements.
A lightweight ComfyUI custom node pack — every node is self-contained with zero dependency on external custom node packs.
Nodes to clean up your workflow.
A Python implementation for integrating the BLIP (Bootstrapping Language-Image Pre-training) model for visual question answering.
Custom ComfyUI node that applies LoRA only during a specific window of the denoising process instead of throughout entire generation.
Runs a shared prompt against a base model and up to 6 LoRAs (with optional master-strength override), tiling the results into one labeled comparison grid.
Bring Microsoft's POML (Prompt Orchestration Markup Language) to your ComfyUI workflows! Create structured, reusable, and powerful prompts with visual node-based editing.
Save images with Adobe RGB / sRGB color space and 8/16-bit depth control for ComfyUI.
ZeroPrompt integration for FLUX2-JSON providing deterministic, profile-based procedural text generation for structured prompt building.
ComfyUI nodes for StreamDiffusionV2 which is an open-source interactive diffusion pipeline for real-time streaming applications
Future-aware smart resident scheduler for ComfyUI low VRAM (--lowvram / --enable-dynamic-vram). Reduces model swap thrashing on 12GB GPUs such as RTX 3060.
A patched version of ComfyUI-NAG implementing Normalized Attention Guidance (NAG) for diffusion models with compatibility fixes for the latest ComfyUI version.
[a/ImageMagick](https://imagemagick.org/index.php) nodes for ComfyUI. Adds nodes to call ImageMagick subprocesses from ComfyUI. Requirements: [a/ImagMagick7](https://imagemagick.org/script/download.php), 'magick' command in your CLI environment.
A ComfyUI output node that saves a cover and content image as an APNG. 输出一张apng,实现在QQ中查看原图才显示真正内容喵
Use VQGAN+CLip
This provides a single node openai > Open AI query node that takes a system prompt and user message and sends them to chatGPT 3.5 Note, you MUST have an OPEN AI API key stored in the environment variable OPENAI_API_KEY in order for this to work.
Text and data manipulation nodes for ComfyUI, with emphasis on JSON processing, detection workflows, and bbox visualization.
Retro graphics converter node for ComfyUI. Converts modern images to authentic retro computing styles (VGA, EGA, CGA, PC-98) with proper color palettes, dithering, and resolution constraints.
Custom nodes with split, random, and select functions for easy visual and management of multiple cumbersome prompts
A premium resolution utility node for ComfyUI with aspect ratio grid, preset selection, custom ratios, favorites, recents, and keyboard navigation.
An interactive JSON prompt editor for ComfyUI optimized for processing visual language model outputs like Qwen3-VQA, with auto-display, local file loading, and export capabilities. (Description by CC)