ComfyUI Custom Extensions
Browse 9,446 ComfyUI extensions and 73,215 custom nodes.
Load Image With FileName
Save and restore widget parameters across ComfyUI nodes with per-node snapshot management and sidebar panel.
Text to animated GIF through the AI GIF Maker API (native GIF generation, not a converted video). Outputs every frame as an IMAGE batch plus the saved .gif. No local GPU; needs an AI GIF Maker API key.
Seedance 2.5 text to video and image to video through the emaki API (Seedance 2.0, MiniMax H3, Veo 3.1 and Nano Banana 2 for first frames too). No local GPU; needs an emaki API key.
Text to video and image to video with sound (MiniMax H3) through the Free Image to Video API: connect any IMAGE as the first frame or leave it empty for text to video; the MP4 is saved to the output folder. No local GPU; needs a Free Image to Video API key.
Generate and edit images through the kavel.ai API. Works without an API key on the free tier; add a key to use your account's models (Nano Banana 2, GPT Image 2, Seedream).
Text to video and photo to video with sound via Movoa (movoa.ai). Standard library only; requires a Movoa API key.
Generate and edit images and animate them into video through the SayMaker API (Veo 3.1, Kling 3.0, Seedance 2.0, Nano Banana 2, GPT Image 2.5). No local GPU; needs a SayMaker API key.
Generate and edit images through the Veida AI API. Works without an API key on the free tier; add a key to use your account's models (GPT Image 2.5, GPT Image 2, Nano Banana 2, Seedream 5.0).
Korean-to-English translator node using Qwen3 model running locally without internet, optimizing Korean prompts for AI image generation. (Description by CC)
NODES: Load llm, Generate Text with LLM, Inference Qwen2VL, Inference Qwen2
comfy ui custom node that returns an image from a batch based on selected criteria such as RGB value, brightness, etc (credits to chris goringe's custom nodes tutorial ).
use case of llm, image, video models on amazon bedrock
A custom ComfyUI node that implements advanced document scanning with automatic perspective correction and multiple enhancement options.
Keyed, iterative and composable low-rank adapters for diffusion models
Starter scaffold for developing ComfyUI custom nodes.
A bounded multimodal OpenRouter node for ComfyUI with text-only output.
A simple node to strip metadata / exif data from image output
Custom nodes to manipulate images in ComfyUI
ComfyUI custom nodes for native LTX Video image guide workflows.
Smart Prompt Manager custom node with managed Helto privacy.
Custom ComfyUI nodes for WAN 2.2 image-to-video guide workflows with multi-image guide management and WAN-native conditioning.
Licensed encrypted LoRA loader for ALTERA2ION ComfyUI packages.
ZEval is a Python library and ComfyUI custom node that allows you to perform mathematical operations and transformations on latent tensors using a domain-specific language (DSL).
Mask-guided precision merge plugin for ComfyUI. Fixes skin misalignment and edge artifacts in crop-and-inpaint workflows by only replacing pixels inside the mask region.
Multiple image upload and loading nodes for handling batch image input in ComfyUI workflows. (Description by CC)
A collection of custom nodes for ComfyUI that provides useful image processing tools.
ComfyUI custom node to convert Japanese text in prompts to Hepburn romaji for LTX-2.3 speech generation
EMA Lightning for ComfyUI: small and fast Turkish text-to-speech (8.6M parameters, 48 kHz, numbers and dates read out, word timings), run in-process through the official ema-lightning package. Bit-identical to the official say() on the GPU.
EmbeddingGemma 2 for ComfyUI: Google DeepMind's multimodal embedding model (text, images, video, audio in one 768-d space) for captioning by similarity, ranking generated images and video or audio search, run in-process through sentence-transformers. Bit-identical to the official usage.
JoyAI-Video-Edit (streaming instruction video editing) for ComfyUI on an RTX 5090: runs the upstream model in a persistent WSL2 worker with a low host-RAM weight placement. Unofficial.
Unofficial community integration, not affiliated with the LeapTalk authors: turn a portrait and speech into a talking video locally in ComfyUI with the official LeapTalk weights and 1-step recipe (SoulX-FlashHead base), run in a separate Python environment.
Unofficial community integration, not affiliated with NVIDIA: applies the NVlabs LongLive-Plug Few-Step/CFG LoRA adapters to Wan2.1-T2V-14B in ComfyUI with key-coverage checks and upstream-verified 4-step sampling recipes.
Looking Ahead (TencentARC, Future Chunk Prediction) 1.3B text-and-image-to-video for ComfyUI: the official autoregressive KV-cache sampler, verified bit-exact against the official pipeline, with ComfyUI's own umT5 text encoder and Wan 2.1 VAE. Unofficial.
Looped-DiT text-to-image for ComfyUI: official B/16 sampling reproduced bit-exactly, plus a loop-depth sweep (same prompt, seed and noise at loops 1/2/4).
mimetic-a0 for ComfyUI: text to music trained only on CC0 and synthetic audio (MIT). Drum loops, basslines, chords, melodies and one-shots at the tempo, key and chords you ask for; 48 kHz stereo, up to 16 s. Runs on ComfyUI's own stable-audio classes with the official sampler: no extra Python packages.
Unofficial ComfyUI nodes for Mi-Ripple's local, deterministic diagnosis and repair of periodic lattice / granule artifacts (CPU, no network, no regeneration). Outcomes follow the official code; verified for 512x512 RGB.
Unofficial community integration, not affiliated with the MonarchRT authors: runs training-free MonarchRT attention on the public Self-Forcing (Wan2.1-T2V-1.3B) weights in a separate WSL2/Linux runtime, with a dense baseline for A/B comparison, an optional persistent worker that keeps the model and tuned kernels loaded between queue jobs, real VIDEO outputs and dispatch/timing reports.
ComfyUI custom nodes for NVIDIA CMD causal image-to-video on Windows Native
Scale-RAE for ComfyUI: text-to-image with the official Scale-RAE model (Qwen2.5-1.5B + 2.4B rectified-flow DiT in WebSSL representation space, RAE decoder), run in-process. Matches the official code bit for bit in the same environment.
SyncReward for ComfyUI: score how well a video's sound is synchronized with its picture (0 to 2), and pick the best of several generated videos. The official model, run in-process.
Unofficial community integration, not affiliated with the TBDub authors: redub an existing video with new speech locally in ComfyUI (TBDub V1.1 Student, 2 steps, MediaPipe face crop and paste-back), run in a separate Python environment.
Standalone ComfyUI nodes for deterministic booru tag normalization against a local Danbooru snapshot.
A simple ComfyUI custom node for creating looping GIFs from VIDEO inputs.
This node provides a button that, when clicked, triggers the execution of the node and generates a new random integer between 0 and 1000. The generated number is available as an output.
ComfyUI custom node that force-aligns audio with text into an SRT file using Meta's MMS models.
A collection of custom ComfyUI nodes for advanced video processing and editing workflows.
Generic ComfyUI custom node for turning a character video into a video-mouth PNGTuber asset bundle.
Unofficial ComfyUI prompt enhancer using Brave Leo through local browser automation.
Qwen multi-angle workflow 向けの白背景自動検出 ComfyUI カスタムノード。画像の四辺エッジをサンプリングして白背景かどうかを判定し、背景プロンプト文字列を出力する。
comfyui node
token validation probe (same version, no-op)
ComfyUI integration for LM Studio language models. (Description by CC)
ComfyUI custom node set that reads .env files at startup and provides dropdown selection or dynamic multi-port output for environment variable access in workflows. (Description by CC)
Jingyuma Storyboard is a ComfyUI custom node that automatically splits story text into shots and generates image prompts, video prompts and dialogue for each, so you can quickly build a storyboard workflow.
LTX Director Enhanced — keyframe timeline + prompt relay for multi-scene LTX 2.5 videos, with in-node QwenVL prompt enhancement. Fork of WhatDreamsCost's LTX Director.
Huelake's ComfyUI custom nodes: XY Grid tools, axis converters, text scheduling, and utility nodes
A ComfyUI custom node for downloading and backing up all custom nodes/plugins as ZIP files.
A ComfyUI custom node for downloading and managing UNet/Diffusion models (FLUX, SD3, etc.) directly within the UI.
A ComfyUI plugin library based on [a/https://github.com/stavsap/comfyui-ollama](https://github.com/stavsap/comfyui-ollama), with the Ollama cluster provided by Huixingyun.
Some custom nodes used for pimg (a comfyui controller deployed in huixingyun).
Simple JS application based on ComfyUI which takes prompt and style picture from user and runs hardcoded workflow inference returning generated image to user.
This is a ComfyUI custom node designed to automatically detect and remove text from images.
Auto-fixes WanVideoModel for PyTorch 2.10.0+ compatibility by automatically patching the model class to work with both PyTorch 2.9.0 and 2.10.0+.
A small ComfyUI custom node that reads txt files from a folder and outputs one file's text.
NODES: Extract JSON, 'Prompt: Render Template from JSON'
A ComfyUI custom node that randomly selects a (Sampler, Scheduler, Steps) combination from user-defined presets.
ComfyUI nodes for Automated Workflow
Custom nodes for ComfyUI that support NP-LoRA and SSR-Merge based LoRA merging, providing loaders for subject-style LoRA fusion, calibration, and merged LoRA application. (Description by CC)
Procedural Multi-character framework & character library with smart-injection technology.
Adds nodes that are designed to be used for scripting ComfyUI. These nodes just describe themselves as 'api'. This can be used in conjunction with the npm library '@hypnodes/comfyui-api-nodes' for scripting via javascript.
Custom ComfyUI node for RunningHub that generates MP4 video from template and overlays handwriting-style text on the red paper area.
Directional Canvas Placement & Outpainting Tools for ComfyUI
One node, instant batch diversity. Same prompt, wildly different images.
Interactive grid cutscene system for ComfyUI that creates polygon masks by clicking points on a grid and outputs masks and previews. (Description by CC)
Includes a CSV parser.
Magik Alert
Magik Basic Nodes
Magik Booleangate
Magik Clone Output
Magik Color Correct
Magik Color Postpro
Magik Delay
Magik Dropdown
ComfyUI nodes for [Animon Platform](https://platform.animon.ai/), a comprehensive API service offering a variety of image and video generation models.
Custom ComfyUI node for running LLMs via HuggingFace pipeline. Supports both local paths and HuggingFace model names.
ComfyUI custom nodes for SoulX-FlashHead — generate talking head videos from a reference image and driving audio.
A ComfyUI custom node for robust human pose estimation. (Description by CC)
Download the model from huggingface and put it in any directory.
Use Claude to expand a short description into a detailed image prompt.
Small image editing canvas tools for ComfyUI. Image Collage: drop, paste or load images onto the node, arrange them in a grid or freely (move, resize, rotate), draw on them, paint an inpaint mask, and output the composited IMAGE and MASK.
A metadata editor that allows add download url and reset widget values
A ComfyUI custom node that resizes images and masks with multiple modes, AI upscaling, various interpolation methods, and live preview.
A collection of custom nodes for ComfyUI
Nodes:LoadFromS3, LoadBatchFromS3, SaveToS3, SaveBatchToS3
Preview-driven execution and sampling trace inspector for ComfyUI
Composable Krea2 guidance and acceleration nodes, including GAG, NAG, Prompt Reinjection, TaylorSeer, and adaptive progressive sampling.
A ComfyUI custom node pack by Lumi.
ComfyUI custom nodes for analyzing mask complexity and routing between direct, simplified, and overlay strategies.
ComfyUI custom node package with a Style Mixer sidebar panel and nodes for managing reusable prompt styles and named mixes.