ComfyUI Custom Extensions

Professional audio processing and mastering suite for ComfyUI.

A ComfyUI custom node for MiniCPM vision-language models, enabling high-quality image captioning and analysis.

ComfyUI MuseV

Save as AVIF, WebP, JPEG, customize the folder, sub-folders, and filenames of your images!

This is a custom node that lets you use Convolutional Reconstruction Models right from ComfyUI. [a/CRM](https://ml.cs.tsinghua.edu.cn/~zhengyi/CRM/) is a high-fidelity feed-forward single image-to-3D generative model.

This is a simple implementation StreamDiffusion(A Pipeline-Level Solution for Real-Time Interactive Generation) for ComfyUI

Better TAESD previews, BlehHyperTile.

Qwen3 TTS for ComfyUI - High-quality text-to-speech with voice cloning, voice design, and custom voices

Kytra's MatAnyone (Video Matting) implementation for ComfyUI - Based on pq-yang/MatAnyone

Enhanced features with flexible choice of inputs and outputs, fine control for pose plotting, freedom to composite poses and fast local pose editting.

Blazingly Fast and Embarrassingly Simple End-to-End Full-Length Song Generation. A node for ComfyUI.

This extension provides various utility nodes. Inputs(prompt, styles, dynamic, merger, ...), Outputs(style pile), Dashboard(selectors, loader, switch, ...), Networks(LORA, Embedding, Hypernetwork), Visuals(visual selectors, )

ComfyUI-FreeMemory is a custom node extension for ComfyUI that provides advanced memory management capabilities within your image generation workflows. It aims to help prevent out-of-memory errors and optimize resource usage during complex operations.

Custom Node for comfyUI for virtual lighting based on normal map. You can use normal maps to add virtual lighting effects to your images.

Scan workflows for required models and download them automatically from HuggingFace, CivitAI, and other sources. Features include filter dropdowns, Tavily AI-powered advanced search, metadata caching, and automatic URL restoration.

Multi-reference and scheduled SCAIL-2 workflow helpers for ComfyUI.

ComfyUI wrapper nodes to use the Diffusers implementation of BrushNet

Encapsulate the commonly used functions of FFmpeg into ComfyUI nodes, making it convenient for users to perform various video processing tasks within ComfyUI.

This repository provides a custom ComfyUI node for running object detection with the [a/Qwen 2.5 VL](https://github.com/QwenLM/Qwen2.5-VL) model. The node downloads the selected model on demand, runs a detection prompt and outputs bounding boxes that can be used with segmentation nodes such as [a/SAM2](https://github.com/kijai/ComfyUI-segment-anything-2).

FL AceStep Training - LoRA Training nodes for ACE-Step 1.5 in ComfyUI

FL CosyVoice3 - Advanced Text-to-Speech nodes for ComfyUI. Features zero-shot voice cloning, cross-lingual synthesis, instruction-based control, and voice conversion using the CosyVoice3 model family. Supports 9 languages and 18+ Chinese dialects with automatic model downloading and caching.

Divide and Conquer Node Suite: It calculates the optimal upscale resolution and seamlessly divides the image into tiles, ready for individual processing using your preferred workflow. After processing, the tiles are seamlessly merged into a larger image, offering sharper and more detailed visuals.

Custom nodes to improve flow control and logic + several utilities to enhance capabilities

ComfyUI custom node of OmniGen project.

Better user experience plugin for ComfyUI.

ComfyUI nodes for ID-LoRA-2.3 one-stage audio+video generation with speaker identity transfer

The official ComfyUI version of facechain greatly improves the speed of reasoning and has great custom process controls.

A ComfyUI workflow customization by Jake.

CRT-Nodes is a collection of custom nodes for ComfyUI

A project for pose alignment and human body proportion mapping based on SDPose.

A comfyui node that provides translation and image reverse push functions(JoyTag & JoyCaption).

ComfyUI Qwen2-VL wrapper that supports text-based and single-image queries.

Instantly explore Civitai community hits for your local AI models and apply full recipes with one click — or dive deep to uncover the best prompts, parameters, and LoRA combos.

All-in-one Civitai integration center for ComfyUI — browse online models, manage local assets,analyze community trends, and instantly apply full recipes within your workflow.

Nodes:LCMScheduler, SamplerLCMAlternative, SamplerLCMCycle. ComfyUI Custom Sampler nodes that add a new improved LCM sampler functions

Extension to show random cat GIFs while queueing prompt.

Everyday utility nodes for ComfyUI: inpaint crop & stitch, upright face crop & stitch with automatic masks, a live face rig for portrait expressions, hand-drawn vector masks, a paint canvas for scribbles and ControlNet guides, path and field blur, mask painting and cleanup, colour grading, gradients, noise and grain — each with a live preview in the node.

Nodes: BilboX's PromptGeek Photo Prompt. This provides a convenient way to compose photorealistic prompts into ComfyUI. Post-Processing: adds various post processing effects. Bonus: Option to show a distant server shutdown menu.

Nodes: Checkpoint Loader with Name, Save Prompt Info, Outpaint to Image, CLIP Positive-Negative, SDXL Quick Empty Latent, Empty Latent by Ratio, Time String, SDXL Steps, SDXL Resolutions ...

Unofficial implementation of [a/UltraEdit](https://github.com/HaozheZhao/UltraEdit) (Diffusers) for ComfyUI

Build a person once, then shoot a whole series: framing, pose, placement, expression and aspect ratio vary while the person stays the same.

Implement Region Attention for Flux model. Add node RegionAttention that takes a regions - mask + condition, mask could be set from comfyui masks or bbox in FluxRegionBBOX node. This code is not optimized and has a memory leak. If you caught a OOM just try run a query againg - works on my RTX3080. For generation it uses a usual prompt that have influence to all picture and a regions that have their own prompts. Base prompt good for setup background and style of image. This is train-free technique and results not always stable - sometimes need to try several seeds or change prompt.

A ComfyUI extension for chatting with your images. Runs on your own system, no external services used, no filter. Uses the [a/LLaVA multimodal LLM](https://llava-vl.github.io/) so you can give instructions or ask questions in natural language. It's maybe as smart as GPT3.5, and it can see.

The Ultimate Local File Manager for Images, Videos, and Audio in ComfyUI

Nodes:Music Gen, Audio Play, Stable Audio

Nodes: Download the weights of MotionCtrl [a/motionctrl.pth](https://huggingface.co/TencentARC/MotionCtrl/blob/main/motionctrl.pth) and put it to ComfyUI/models/checkpoints

FeiHou Toolbox: API image/video generation, standalone multimodal media loading, reference-image, mask, seed-noise, and one-file VHS video saving tools.

ComfyUI_RH_OminiControl is a ComfyUI plugin based on OminiControl By splitting the pipeline load, the plugin efficiently runs on NVIDIA RTX 4090 GPUs. Additionally, the spatial and fill functionalities are generated using the schnell model, reducing the number of sampling steps and improving overall efficiency.

Additional UI focused on inference. Stable UI states; presets; and advanced queue. Based on Gradio

Warning, uses experimental package `comfy-env` to attempt a one click isolated install. Will download and use pixi package manager. ComfyUI wrapper for LiTo: single-image to 3D Gaussian Splat generation

A universal swap node that supports ComfyUI native workflow, allowing 4_6G users to experience Klein9B or other large models

An execution-scoped F1B0 residual block cache for ComfyUI's native MiniMax H3 model.

Simple custom nodes for testing and use HiDiffusion technology: https://github.com/megvii-research/HiDiffusion/

Use of the molmo model.Generate detailed image descriptions and analysis using Molmo models in ComfyUI.

ComfyUI Flux Accelerator is a custom node for ComfyUI that accelerates Flux.1 image generation, just by using this node.

ComfyUI implementation for [a/TCD](https://github.com/jabir-zheng/TCD).

Runware Inference API Integration for ComfyUI (No GPU Required).

DaSiWa Custom Nodes Collection

An easy-to-use ComfyUI plugin for LLM integration supporting DeepSeek, OpenAI API-compatible models, with video reverse-prompt (captioning) capabilities. (Description by CC)

InfiniteTalk lip-sync node optimized for Wan2.2 dual-model workflow with first/last frame precision control and FPS synchronization. (Description by CC)

ComfyUI custom nodes for LongCat-AudioDiT TTS - zero-shot and voice cloning diffusion TTS

Nodes:LoadImageWithSwitch, ImageBatchOneOrMore, GenderControlOutput, ImageCompositeMaskedWithSwitch, ImageCompositeMaskedOneByOne, ColorCorrectOfUtils, SplitMask, MaskFastGrow, CheckpointLoaderSimpleWithSwitch, ImageResizeTo8x, MatchImageRatioToPreset, MaskFromFaceModel, MaskCoverFourCorners, DetectorForNSFW, DeepfaceAnalyzeFaceAttributes, VolcanoOutpainting, VolcanoImageEdit, ReplicateRequstNode etc.

powerful nodes for wan2.1 vace

Proper implementation of ImageMagick - the famous software suite for editing and manipulating digital images to ComfyUI using [a/wandpy](https://github.com/emcconville/wand). NOTE: You need to install ImageMagick, manually.

Run FP8 and INT8 quantized models (FLUX, SD3.5, Ideogram 4, Krea2) on Apple Silicon / MPS — compatibility fixes plus bit-exact fp8/int8 Metal matmul kernels, plus a psutil fix for recent macOS betas.

A ComfyUI node to automatically extract masks for body regions and clothing/fashion items. Made with 💚 by the CozyMantis squad.

ComfyUI node that pixelizes images.

Use llama.cpp to help generate some nodes for prompt word related work

AI Tool Use API for Anima anime/illustration image generation. Supports MCP Server (native image display in Cursor/Claude), HTTP API, and CLI.

Click-to-refine sampler. Generate, then click or paint regions to refine them in place. First-class support for FLUX 2 Klein 9B and Qwen-Image-Edit (Smart Inpaint / Smart Guided Inpaint / Xtra-Fine), works with any sampler-compatible model for Refine + Area Prompt. Optional SAM 3 Detect for text-prompted segmentation, and a companion Overrides node for custom samplers (power-sigma, Flux 2 scheduler, NAG).

Random nodes for ComfyUI I made to solve my struggle with ComfyUI (ex: pipe, process). Have varying quality.

A ComfyUI custom node for Untwisting RoPE.

ComfyUI plugin that integrates Pascal 3D architectural editor

An advanced Assets manager for ComfyUI outputs with gallery, metadata inspection, ratings, and tags.

Powerful ComfyUI custom node built on the FlashVSR model, facilitating real-time diffusion-based video super-resolution for streaming applications.

A set of custom nodes that pause the flow to allow you to pick images, edit parameters, set masks etc..

FSampler (fsampler) is a training-free, sampler-agnostic acceleration layer for diffusion sampling that reduces model calls by predicting each step's epsilon (noise) from recent real calls. Provides fixed history modes (h2/h3/h4) and adaptive mode for 40-60%+ speedup.

Powerful node for generating long-form videos with consistent motion, global scene coherence, and slow-motion correction in Wan 2.2-based workflows.

This extension provides a ComfyUI Custom Node implementation of the [a/Depth-Anything-Tensorrt](https://github.com/spacewalk01/depth-anything-tensorrt) in Python for ultra fast depth map generation

FL HeartMuLa - Multilingual AI music generation nodes for ComfyUI. Generate full songs with lyrics using the HeartMuLa model family. Supports English, Chinese, Japanese, Korean, and Spanish with song structure control via section markers and style tags.

Describe a single image or all images in a directory using models such as Janus Pro, Florence2, or JoyCaption (testing), with a particular focus on building datasets for training LoRA.

A Flux2 text-to-image and image-to-image editing node supporting dual modes, multi-image editing, intelligent masking, and flexible resolution settings. (Description by CC)

ComfyUI custom nodes for Woosh Sound Effect Foundation Model by Sony AI — text-to-audio, video-to-audio generation

A lightweight ComfyUI custom scheduler & sigma generator for Z-Image-Turbo. Delivers a stable linear sigma schedule (1.0 → 0.0) for rectified-flow/flow-matching, boosting few-step generation (8-9 steps) with superior consistency, reduced noise, and full compatibility with the model's distilled pipeline.

Collection of utility nodes for ComfyUI designed specifically for the Z-Image model with vision model support and LLM-powered prompt enhancement.

Towards GPT-4 like large language and visual assistant.

ComfyUI GLM Nodes: Seamlessly integrate ZhipuAI GLM models into ComfyUI for enhanced text generation and image-to-prompt capabilities. Features include advanced text chat (video prompt expansion) and GLM-4V powered image description.

Simple use of [a/Mflux](https://github.com/filipstrand/mflux) in ComfyUI, suitable for users who are not familiar with terminal usage. NOTE: A MLX port of FLUX based on the Huggingface Diffusers implementation.

Text translation node for ComfyUI: No need to apply for a translation API key, just use it. Currently supports more than thirty translation platforms.

GrsAI API node supports models: Flux-Pro-1.1 (¥ 0.03), Flux-Ultra-1.1 (¥ 0.04), Flux Kontext Pro (¥ 0.035), Flux Kontext Max (¥ 0.07), GPT Image (¥ 0.02). Support text generated images, image generated images, and multi image fusion.

Better Dynamic, Higher Resolution, and Stronger Coherence!

This node allows downloading models directly within ComfyUI for easier use and integration.

A ComfyUI extension providing nodes for generating 3D models from images and rendering/visualizing 3D assets within your workflow.

ComfyUI custom nodes for NVIDIA PiD pixel diffusion decoding with VRAM offload support.

This is the rgb2x wrapper node for ComfyUI. The required models are automatically downloaded on the first run. original project : [a/https://github.com/zheng95z/rgbx](original project : https://github.com/zheng95z/rgbx)

Build Live AI Video with ComfyUI

This is a ComfyUI plugin for [a/HYPIR (Harnessing Diffusion-Yielded Score Priors for Image Restoration)](https://github.com/XPixelGroup/HYPIR), a state-of-the-art image restoration model based on Stable Diffusion 2.1.

Implements iteration over sequences within a single workflow run. [w/NOTE: This node replaces the execution of ComfyUI for iterative processing functionality.]

Multi-artist mixing for Anima through cross-attention or post-adapter embedding projection.