ComfyUI Custom Extensions
Browse 9,446 ComfyUI extensions and 73,215 custom nodes.
ComfyUI custom nodes for Pipe-based workflow management, KSampler, Detailer, Upscaler, SAM3 segmentation, Toggle Manager, and Collector nodes.
ComfyUI custom nodes for BespokeAI image-to-3D generation NOTE: The files in the repo are not organized.
Utility custom nodes for ComfyUI: prompts, resolution, safe math, seeds, precision-aware random numbers and YuE2 instrumental preparation.
Node for using the out-of-tree torchvulkan backend with comfyui
ComfyUI custom node for flexible image loading with file picker and random folder selection. Features instant preview, auto mode switching, and browser-based file selection.
Generate and edit images using Google Gemini models in ComfyUI with ComfyUI Cloud API authentication.
A Windows 95 style 3D dungeon crawler generated by your ComfyUI - served at /comfycrawler/ with a button in ComfyUI's menu. Adds no graph nodes.
Anima Block Compile for ComfyUI: a one-purpose node that runs torch.compile on the Anima DiT per transformer block (diffusion_model.blocks.{i}) instead of the whole model. Faster compile, fewer graph breaks. A thin, Anima-named wrapper over ComfyUI core's set_torch_compile_wrapper.
Anima loader nodes for ComfyUI. AnimaAdapterLoader auto-detects LoRA, HydraLoRA multi-head with live σ-conditional and FeRA-style FEI-conditional routing, ChimeraHydra dual-pool routing (per-Linear content router + network-level FreqRouter on FEI+σ), and LoReFT residual-stream edits. AnimaFeraLoader handles FeRA (Yin et al., arXiv:2511.17979) — both the author-faithful networks.methods.fera format and the plan2 stacked_experts_global_fei format with global router on the latent's spectral energy + independent stacked experts per Linear. AnimaSoftTokensLoader handles SoftREPA-parameterization soft tokens (Lee et al., arXiv:2503.08250) — per-layer × per-timestep-bucket learned tokens spliced into the crossattn embedding inside the first n_layers DiT blocks. Three single-purpose nodes; chain via the MODEL socket.
Decode Anima renders straight from the DiT's last-layer features with a 25M-parameter pixel head, in place of the VAE decode: a MODEL tap (stock or Spectrum KSampler) and a decode node that outputs IMAGE.
A custom node that adds checkpoint display with trigger word descriptions loaded from checkpoint_triggers.txt file. (Description by CC)
A custom ComfyUI node for downloading YouTube videos and audio with precise resolution control
This is a node that simply integrates LoadImage, Vae Encode, Upscale, Resolution factor correction, and Color Adjustment.
ComfyUI-LoadLoraModelOnlyWithUrl
Load-time patch so ComfyUI loads depth-expanded Anima models (Anima-2.9B, 40 blocks) correctly instead of silently truncating them to 28 blocks.
Frontend-only extension that filters COMBO dropdown values with ordered whitelist and blacklist rules.
Node for scheduling conditioning strength while preserving non-floating tensors such as token ids.
Suppresses console messages from JavaScript to improve performance.
Lazy indexed switches with a default route for ComfyUI.
RulZ nodes for ComfyUI: Load Image with HEIC/HEIF support and a filename output, a resumable one-image-per-run batch loader, a Reference loader with 1 to 16 image slots and outputs, an Image Browser sidebar tab that sends selected images from any folder to the batch loader, Mask Intelligent Crop, an advanced Resize Image with megapixel and input-size targets, DeskinTexture to remove honeycomb skin texture, a per-workflow Disable Comfy Compiler model patch, a seeded Random Prompt Selector that picks a line from a text file, a click-to-mask Person Select built on SAM 3, block-sparse SageAttention for Qwen Image 2.1, and a Generate Text that writes several answers in one pass.
Video frame picker + trimmer for ComfyUI. Scrub, pick FF/LF, pipe into first-frame/last-frame video models.
Compact, UI-friendly nodes for ComfyUI with LoRA/checkpoint selectors with thumbnails, text utilities, image helpers, and a lightweight cache cleaner script.
A set of nodes developed from the original Segment Anything set, to streamline and fix new library conflicts
Adds a StableYogi Prompt node that outputs a ready-made prompt string. Wire it into CLIP Text Encode and every queue pulls a fresh prompt.
allowing the modified seed to truly make a difference
The ComfyUI node version translated by HY-MT1.5
Prompt cycling with auto prompt count and auto queue; SeedVR2 per-image folder upscaling saved next to the source (with native folder/image picker buttons), scale/long-edge settings, jpg/png/webp output settings, and a Save Image node that follows the output settings. Writes a per-date 还原记录.csv (one row per saved image: file, real prompt, seed, models, LoRAs, sampler settings).
ComfyUI nodes for saving, loading, and versioning GenAsset generations.
Concatenate multiple text inputs into a single output with auto-expanding inputs and configurable separators.
A custom node for ComfyUI that cycles through a list of up to 10 image inputs in sequence on each generation, with automatic wrap-around and a reset toggle.
Enhanced fork of ComfyUI-Kimodo — text-driven 3D human motion (NVIDIA Kimodo) with a rebuilt retarget/export stack: glb + FBX export, A-pose arm fix, travelling root motion, finger and orientation handling.
Custom nodes for ComfyUI including LLM Chat (Claude, Gemini, GPT), WaveSpeed AI image and video generation, VertexAI integration, and workflow utilities.
A ComfyUI node for image layering, blending, compositing, and batch processing.
Fork of Ultimate Lora Loader with folder browser, Civitai names/covers, and Toolkit-style info buttons. Runs standalone, but for FULL features install TWO other custom_nodes separately: Civitai Toolkit and ComfyUI-Custom-Scripts (pysssss). Tested Toolkit 4.1.3 + Custom-Scripts 1.2.5.
Local prompt combine for ComfyUI: named group cards → str_pos/str_neg, External wired slot, SQLite presets (structure) / templates (full packs) / pairs, sidebar pin, manager, optional autocomplete + per-node CLIP PNG embed.
A ComfyUI custom node that loads character definition YAML files and lets you compose prompts via per-category dropdowns with live rescan support.
ONNX (FP16 + CUDA) で wd-eva02-large-tagger-v3 を呼び、タグ文字列と検閲状態 4 値 (uncensored / censored / no_explicit / unknown) を同時に返す ComfyUI ノード。
Local PLaMo 3 Translate GGUF translation with CPU/CUDA modes and bounded GPU model reuse.
ComfyUI custom node that loads UNET diffusion models and extracts the model name as a string output.
Nodes:YOLO Classifier Model Loader, YOLO Classify.
Custom ComfyUI nodes wrapping fal.ai models for image editing, image-to-video, upscaling, vision/language, and region selection - built for Superside production workflows.
MiniMax H3 智能参考标签提示词助手,附带 Ares 视频工具节点
暂停工作流进行人工文本编辑的 ComfyUI 节点
Visual layout engine for ComfyUI: turns shot definitions into composition layouts and normalized bounding boxes as editable JSON. Consumes shots; does not generate them. Not a prompt or image generator.
已改名为 SnackNodes
Save ComfyUI outputs to your own Google Drive, Cloudflare R2 (10 GB free) or WebDAV server (Nutstore, Nextcloud, NAS...), connected right on the node, optionally end-to-end encrypted as a Cryptomator vault.
A ComfyUI node that combines an image with simple text parameters to create a prompt, sends it to the Google Gemini API via the google-generativeai SDK, and returns the generated text response along with the original prompt and an execution log
Advanced custom node for ComfyUI with an integrated sketching and mask drawing editor built directly into the node interface for drawing masks, sketches, and generating Ideogram prompts.
NODES:Sticker Compositer. background frame + sticker
Sidebar browser for SWORKS_TEAM LoRAs from Civitai. Browse, download, and apply LoRAs with image previews and tag filtering.
A Blender add-on that integrates with ComfyUI for AI-powered model generation. Provides workflow management, image input handling, and real-time status tracking for seamless AI art creation within Blender's interface.
A node that displays any node's history
MiniMax H3 text-to-music / audio-only pipeline via shrunk video canvas
Fibo 2.0 Dev and Edit Turbo official pipelines, separate DiT/VAE/TE loaders, and Comfy-Org model support.
WeVisOne visual perception, Mask2Alpha, geometry and video tools for ComfyUI
TK View and Light: A ComfyUI node for 3D camera and lighting control
Unofficial secure OpenRouter image and video generation for ComfyUI, with model-aware controls and resumable jobs.
Turn one image into a camera you can fly: a live 3D Gaussian-splat viewport inside a ComfyUI node, with mask, depth, and orbit recording for camera-motion reference.
Intelligent subfolder auto-linker and model replacement suggestion assistant for ComfyUI.
Provides stable pitch shifting in ComfyUI, optimized for mono processing and seamless MMAudio integration, using a pure CPU phase-vocoder time-stretch to control audio pitch.
A specialized suite of custom nodes for ComfyUI designed to hunt down and visualize generation parameters (prompts, steps, samplers, etc.) embedded in AI-generated images. It works across different formats and platforms (ComfyUI, Automatic1111, Civitai, and more).
A Fooocus-style web interface on top of ComfyUI: prompt, styles, LoRAs, upscale/variation, inpaint/outpaint and metadata on one page, served at /fooocus. Adds no nodes.
ComfyUI node to use advanced Torchvision transforms built for accuracy, visual quality, and speed.
Unified multimodal captioning node for ComfyUI with OpenRouter and Replicate support.
Comprehensive PyTorch nodes for ComfyUI - Neural network training, inference, and ML workflows
Generate or draw FLOAT arrays within ComfyUI
A ComfyUI node that outputs X/Y floats (0.1..0.9) using a clickable 5x5 grid UI, with an optional Random toggle that randomizes the grid selection each queue.
ComfyUI wrapper nodes for Image Transformer
Save and Reload an Image to prevent first run OOM
A swiss‑army ComfyUI node that lets any vision‑capable LLM (OpenAI GPT‑4o, Gemini 1.5 Pro Vision, Groq‑mix… you name it) rip through a reference image and return a prompt you can feed straight back into your diffusion pipeline—no hand‑crafting required.
a single director node for video workflows and adapters for popular video models
Small workflow navigation helper for large ComfyUI graphs with many grouped workflows, featuring switchboard UI for selecting and managing workflow groups.
ComfyUI custom node that activates integration with a Sentry instance for loading. Has no actual nodes.
Nodes: MPN Switch, MPN Reroute, PN Switch. This is a set of custom nodes for ComfyUI. Mainly focus on control switches.
Compatibility fork of ComfyUI-NAG (Normalized Attention Guidance) that fixes the Chroma layer imports broken by recent ComfyUI core refactors.
RunningHub Nano Banana Pro API node for ComfyUI
This is the ComfyUI extension for use with texflow. It provides two new nodes, 'Load Texflow Depth Image' and 'Save Texflow Image'. For more information see [a/the main texflow repo](https://github.com/theAdamColton/texflow/)
ComfyUI custom nodes for BytePlus / ByteDance Seedance 2 video-generation API.
DIGIT — Deep Infrastructure for Generative Image Training. ComfyUI nodes for Gemini generation, Veo + Seedance 2.0 / 2.5 video, LoRA training, dataset management, and auto-captioning.
This repository contains a collection of custom nodes for ComfyUI designed to integrate external AI models, provide utilities, and enable advanced workflows.
A lightweight ComfyUI custom node for generating structured exhibition booth prompts.
Custom ComfyUI node providing camera angle and cinematic shot selection presets designed for Qwen multi-angle workflows.
Local-first multimodal LLM nodes for ComfyUI (GGUF+mmproj via llama.cpp, optional Transformers fallback).
A ComfyUI node for generating rich, detailed captions for images using NVIDIA's vision models. Supports batch processing, multiple captioning styles, and includes built-in caching for efficient workflows.
A universal node for generating empty latent tensors with support for SD3.5, SDXL, and Flux models. Features extended aspect ratio support, batch processing, and flexible dimension overrides.
Enhanced empty latent node with extended aspect ratio support for ComfyUI
Load Random Text is a ComfyUI node pack that instantly picks a text file from a folder and outputs its contents straight into your workflow.
Custom nodes for ComfyUI implementing the csm model for text-to-speech generation.
20-band LoRA equalizer for ComfyUI that surgically adjusts LoRA influence across frequency-like bands, cross-attention, and MLP layers. (Description by CC)
RIFE frame interpolation for ComfyUI using NVIDIA TensorRT acceleration with dynamic shape support and persistent caching.
NVIDIA TensorRT-accelerated upscaler with dynamic shape support, up to 30x faster upscaling with automatic engine building and persistent timing cache. (Description by CC)
This node allows you to loop through specified layers in the LTX 2.3 model to potentially improve output quality with an eye toward artifacts in high motion.
INT4/INT8 quantized inference for ComfyUI diffusion models
A collection of custom nodes for ComfyUI, designed to provide utility functions for list processing, file management, and more.
For users who merge models in ComfyUI and want the saved checkpoint to match the live merged model exactly. A small custom node for saving a live merged MODEL without using ComfyUI's normal ModelSave materialisation path.
Potential manual fix as a custom node for ComfyUI for users seeing memory bug issues
prediction boost custom node for ComfyUI
ComfyUI custom nodes for Higgs Audio v3 TTS with native inference, voice cloning, multi-speaker dialogue, longform chunking, and AIMDO DynamicVRAM support
Text Randomization and Formatting, JSON Extraction and Processing, SD3 Resolution Solver
Nodes to run ComfyUI workflows in isolated workers for optimal GPU and memory use.