ComfyUI Custom Extensions
Browse 9,446 ComfyUI extensions and 73,215 custom nodes.
A ComfyUI custom node that converts Civitai share/model page URLs into direct download URLs, making it easier to integrate Civitai models into your workflows.
Call the local `claude` CLI from ComfyUI for vision+text or text-only completions. Uses your existing `claude login` OAuth (no API key required). Supports UUID session reuse for prompt-cache savings, with session token-usage tracking. Additional utility nodes: `Split Image Batch to List of Image Batches` (takes a large incoming batch and splits it into a list of smaller sub-batches with start/end index outputs, so you can send N images at a time — e.g. pairs — to the Claude CLI node along with their original batch indices for tracking), `Indexed 2x2 Grid from Batch`, `Select Images (Bounds Safe)`, and `Image Batch Current Index`. Optional headless API-key mode.
A custom node extension for ComfyUI that enables conversion between different prompt formats: NovelAI V4, ComfyUI, and old NovelAI.
Seasonal fashion/scene prompts and local AI prompt writing for ComfyUI, with optional TIPO text/image tag expansion.
LoRA and Checkpoint loader nodes with integrated Civitai metadata, preview images, trigger word management, and a full-featured model manager UI. Features auto-fetch from Civitai API, multi-threaded downloads, retry logic, and structured metadata storage.
Added mask area.
A ComfyUI custom node that enhances image loading with subfolder organization and dynamic filtering
GGUF loader + text-to-image nodes for NVIDIA NL-Diffusion-Image (masked discrete diffusion LM + IBQ VQ decoder) on consumer hardware.
A minimal ComfyUI extension that lets you mark any node to discard its output from cache immediately after all downstream nodes have consumed it.
Browser-based screen capture node enabling real-time region of interest selection and cropping for ComfyUI deployments. (Description by CC)
Official Renest panel: save a ComfyUI run that worked - models, custom nodes, exact package versions, workflow - and bring it back on another Linux GPU machine, every file checked against its SHA-256.
ComfyUI custom node that loads and encodes prompts from TXT files with fixed/increment/decrement/random modes for batch image generation.
Interactive button-based trigger word manager with precision weighting and portable collections.
Restore pseudo-pixel art to genuine pixel art with automatic detection of pixel block side length. 自动检测像素块边长,把伪像素图还原成真正的像素图
A collection of nodes for rerouting multiple I/O lines together in a bus.
RunningHub MiniMax-H3 audio-video custom nodes for ComfyUI (direct, in-process)
Official diffusers Qwen-Image 2.1 pipeline as a ComfyUI node (unified text-to-image / image-conditioned generation) with ComfyUI DynamicVRAM streaming.
Danbooru tags -> captions in the style of the Krea 2 Anime-tuned training captions (no LLM, CPU), with tag autocomplete.
kijai ComfyUI-PromptRelay with fps and max frame output
Manage models: browsing, download and delete. Fork of ComfyUI-Model-Manager with improved HF/Civitai auth, huggingface_hub integration, and bug fixes.
A universal signal notifier node for ComfyUI that plays configurable sounds (voice modes, custom audio) whenever it receives any signal such as image, latent, or conditioning. (Description by CC)
Gemini-powered image generation for ComfyUI supporting text-to-image, image-to-image, and batch processing with flexible API configuration.
A custom node to resize your image to model-specific resolution
Pixel-precise, interactive video cropping and time/frame-range trimming - powered by FFmpeg.
Nodes: RK_CivitAIMetaChecker, RK_CivitAIAddHashes.
Poseable 3D OpenPose skeletons (body, hands, face) with a moveable camera, rendered as a ControlNet pose image.
ASDF-style pixel sorting implementation in Python with extended features
This project is the comfyui implementation of ComfyUI_SZtools, a labeling and naming tool developed for Kontext's local training package T2ITrainer.
A custom node for ComfyUI that applies a Gaussian blur to a mask.
Dynamic prompt generation with {A|B} variants, __wildcard__ files, and ${var} variables
Kokoro TTS's Chinese comfyui node, 100 speakers
A custom node for ComfyUI that provides pose alignment functionality.
A real-time StreamDiffusion pipeline node for ComfyUI with integrated ControlNet, IPAdapter, and LoRA support.
Lightweight custom node for ComfyUI that displays backend queue count during workflow execution with minimal performance impact (0.004s overhead) and flexible display modes. (Description by CC)
ComfyUI string-joining and prompt-composition nodes for combining outputs from ComfyUI-PromptRandomChoice and other STRING sources, with optional inputs, queue-fixed/runtime toggles, and live text updates during queued generation.
A set of ComfyUI nodes designed to enhance your workflow with realistic filename generation and keyword generation.
Offload CLIP / T5 / Gemma text encoding to a separate GPU machine and use the results directly in ComfyUI.
A fully interactive calculator node for ComfyUI with variable inputs and scientific functions
Nodes:ImageLoadFromBase64, ImageLoadByPath, ImageLoadAsMaskByPath, ImageSaveToPath, ImageSaveAsBase64, VHSFileNamesToStrings(IOHelpers).
A small, practical bundle of ComfyUI nodes that streamlines common workflows.
Pose Control LoRA nodes for Krea-2 in ComfyUI
An intuitive prompt previewer ComfyUI node for use with Fal's Multi-View LoRA, optimized for portraits with real-time prompt previews.
Two powerful custom nodes for ComfyUI to create stunning before-and-after transition videos. These nodes are designed for visual comparisons, transformations, and creative effects, supporting both standard and depth map-based transitions.
A ComfyUI custom node for authoring structured JSON prompts (Ideogram-style) with a visual, drag-to-draw bounding-box layout editor and optional VLM/LLM agents.
One-node hires upscale for ComfyUI: draws, model-upscales (ESRGAN-style) and redraws in a single run. Works with Qwen Image 2.1 edit, Krea 2, Z-Image Turbo and other text-to-image models.
Nodes for Video effects. Zoom In/Out - new features soon.
Validated RC2 audio conditioning for LTX-2.5 lip-sync, with separate clean output audio.
This custom node for ComfyUI provides full-body animation capabilities, including facial rigging, various lighting styles, and green screen output.
Frame-based prompt scheduling for InfiniteTalk/MultiTalk in ComfyUI
A native, high-performance signal-routing, workflow-organization, and virtual-memory architecture for ComfyUI
Nodes for YuE2 ABC scores: change its tempo, search for a score that fits a target song length, check a score's length, and repair meter glitches.
Load media (image/video/audio) from remote URL
You can use memeplex and DALL-E thru ComfyUI. You need API keys.
ComfyUI experimental node for observing cross-attention in Anima/Cosmos models during image generation, outputting analysis to JSONL/CSV/Markdown formats. (Description by CC)
This is a custom node to add timestep for FreeU V2.
This is a custom node to add timestep for FreeU V2.
Flexible preset selector with YAML support, advanced keyword filtering, hierarchical key search, multi-file/folder search, and subfolder organization
A ComfyUI custom node for randomly selecting and applying LoRAs from multiple folders with support for trigger words, sample prompts, and strength randomization.
Break the execution, save the incompleted image then continue later.
Extract metadata from image.
Load image from directory.
Modify text by condition.
Best-of-N seed search for video and image models: the N drafts run under native EasyCache, a verifier (ImageReward) scores them, and only the winning seed is re-sampled at full compute (CachedSearch, arXiv 2607.23159).
Prompt parser which applies attention bias instead of weights. Use '<' and '>' to enclose it and '=1.0' to specify attention strength. Example: 'This is <a huge dog=1.25>'.
Simple restart button embedded in ComfyUI menu.
Converts Danbooru-style tag lists to natural English prompts using local LLM, preventing concept bleeding in multi-character generation with smart hybrid execution engine. (Description by CC)
Plugin that implements world automatic typesetting and outputs only one paragraph of text
Garment-centric human generation nodes for ComfyUI using DreamFit with Flux. DreamFit is a powerful adapter system that enhances Flux models with garment-aware generation capabilities, enabling high-quality fashion and clothing generation.
Slapshot nodes for ComfyUI — Rotoscoping, Depth Map, Camera Tracking, and Motion Vectors. Submits video jobs to the Slapshot API and reports completion. Visit https://github.com/slapshot-ai/comfyui-slapshot for details.
Simple ComfyUI custom nodes for generating FLUX.2-style prompts with a local LM Studio model, with simplified controls optimized for cleaner output.
Unofficial Venice AI API nodes: every Venice image, image edit (one image or up to 10 references) and video model (prices and censored/uncensored in the model list, per-model durations/resolutions, cheap preview runs, price limit) plus Video Save Plus, Save Image + Prompt TXT and Load & Resize Image + Filename. Needs your own Venice API key. Not affiliated with Venice.ai.
ComfyUI custom nodes that write prompts for image and video models, with the xAI Grok API or a local uncensored Qwen3.8-27B (LLM GGUF): ideas or captions in any language become optimal English prompts for Krea 2, Qwen-Image 2.1, Z-Image, FLUX, Illustrious, Pony, SDXL, Wan and MiniMax H3, and for the API models Seedream, Seedance, GPT Image, Nano Banana, Grok Imagine, Flux 3, Kling and Veo. Editable task presets with filters, your own persons, groups and style blocks with ready-made template persons and a separate Persons node, prompt archive, up to four reference images, standard negative prompts, LoRA caption refinement and image captioning.
A/B image interactive comparison node for ComfyUI with multiple comparison modes (A, B, Slide). (Description by CC)
A ComfyUI custom node that extends the standard Load Image functionality with a convenient 'Paste from clipboard' button. This allows you to directly paste images from your clipboard without workflow conflicts, ensuring only images are pasted.
SnJake Sapsan-VFI is a custom ComfyUI node for x2 video frame interpolation. It inserts one in-between frame for every input pair, doubling the FPS.
AP-Wildcards is a powerful node for ComfyUI that allows using wildcards (random value substitution) in prompts. Supports hierarchical category organization, search, templates, and complex logical constructs.
Persistent local GGUF LLM server, OpenAI API, generation/settings nodes, and Prompt Enhancer
Flex Attention (torch.compile) node for ComfyUI. Uses PyTorch's built-in flex_attention with torch.compile — no extra dependencies, works on any GPU. Falls back to PyTorch SDPA when masks are present.
Text-to-image prompt enhancer for ComfyUI using an OpenAI-compatible chat endpoint
Run Comfy Router image, video, and audio models from one ComfyUI node with provider selection and cost references.
Virtual Set and Get nodes for ComfyUI: name a value on a Set node and read it from a Get node without long wires.
Allows you to double-click on the thumbnail of a node to open the full-size image in a new tab.
Frontend-only nodes that publish or read many named variables at once: a rack of setters, and a rack of getters.
A node for ComfyUI that modifies the values in the samples input that fall outside of a min and max range using a multiplier.
Deletes or moves files out of the `temp` directory to improve performance. Highly customizable.
Resize an image to the closest standard aspect ratio at a target megapixel count, snapped to a selectable pixel multiple, with image, VAE latent, empty latent (SDXL / Flux.2 Klein) and width/height outputs.
Depth/luma-driven Differential Diffusion mask node for ComfyUI. One node: image in, model + latent with noise mask out.
Group nodes into a clean card UI — a facade over your existing graph.
Side panel companion to Vewd. A persistent right-side tab for reviewing, hearting, and exporting generated media — auto-captures images, video, audio, 3D, and splats from any workflow.
This is a simple but powerful custom node for ComfyUI, designed to help you manage and sequence multiple prompts from a single, clean interface. It's ideal for workflows like sequential video generation where you need to feed a series of different prompts into your model.
ComfyUI model inspector with an interactive architecture report.
Blends two image batches frame-by-frame using vectorised tensor ops. Supports 15 blend modes (normal, add, multiply, screen, overlay, soft light, and more). No loops or workarounds needed — works natively with batches.
The implementation of Comfyui in the Sapiens2 project
ComfyUI nodes for SenseNova U1.5 text-to-image generation and multi-reference image editing
Local HPSv3 and HPSv3++ NF4 image scoring and prompt generation for ComfyUI
Run ByteDance Seedance 2.0 video generation in ComfyUI via Sjinn.ai API. Supports up to 6 image, 3 video and 2 audio references. Outputs native VIDEO plus first frame, last frame and frame batch.