ComfyUI Custom Extensions

Fork of ComfyUI-Inpaint-CropAndStitch adapted for Nano Banana 2. Adds NB2 Mask Generator, exact-resolution cropping and feathered alpha compositing in the stitch step.

VFX I/O & color toolkit for ComfyUI: image + video read/write (OpenImageIO + PyAV), OCIO 2.x color management, Nuke-style Grade + Color Correct, OpenCV reformat, frame-order reverse, render-farm-safe Seed, and frame-range slicing. Native VIDEO sockets on every pixel node for low-RAM streaming. 13 nodes.

ComfyUI custom nodes for the Beeble AI API (SwitchX video-to-video compositing, with room for additional Beeble models).

Local AI prompt generator using Qwen/SmolLM2 models. 100% offline and private. Supports 4-bit/8-bit quantization. Runs on 6GB VRAM GPUs alongside Stable Diffusion. Smart token management, Polish-English translation, embedding support, OOM protection.

An advanced, zero-dependency AI-powered custom node for ComfyUI. Designed for flawless face cropping, zooming, and padding using a local YuNet-based approach.

Render Mermaid, PlantUML, Graphviz, D2 and 28 Kroki diagram types as PNG/SVG/PDF in ComfyUI. Uses the Kroki API (no local setup) or optional local Mermaid/Graphviz. Includes a built-in zoomable viewer (Save to ComfyUI, shareable URLs), optional cairosvg for in-node SVG/PNG preview, and LLM Prompt Engine + LLM Call nodes to generate diagram code from templates.

Antonioilev Light Pack - practical ComfyUI nodes for 3D workflows, mesh repair, UV processing, PBR previews and texture tools.

Visual prompt-library manager for Ideogram 4 (ApoGalleria-Ideo4), plus sibling converters that turn its capt...

Powerful ComfyUI custom node for dynamic prompt generation using wildcards and bracket expressions, enabling infinite variations with weighted selection, multi-picks, and nested expansion support.

Model-aware preset library for high-end architectural visualization. One preset stack, compiled to the native dialect of Nano Banana, Flux 2, GPT Image 2, or Ideogram 4. Three nodes: (AV) Preset, (AV) Matrix, (AV) Assembler.

ComfyUI node that lets the user store indexed collection of images with associated properties.

Workflow-focused ComfyUI extensions for image generation.

Modifies execution to minimize RAM at the cost of performance

Custom ComfyUI nodes for generating prompts using Grok AI, enhancing prompt creation for text-to-image workflows.

All-in-One Quality of Life, Smart Presets Hub, Auto Model Assigner & Visual Regional Prompting Suite for ComfyUI

A collection of helper nodes for working with LLM APIs in ComfyUI, intended to complement other LLM custom nodes.

Frontend-only node to display ComfyUI previews

a plugin of ComfyUI for Long Sonic

Civicomfy seamlessly integrates Civitai's vast model repository directly into ComfyUI, allowing you to search, download, and organize AI models without leaving your workflow.

This extension provides nodes that allow experimentation with various elements (samplers, latent, activators, attenuator, scheulders, ...) of Stable Diffusion.

Shows which models you actually use vs. which are just wasting disk space. Read-only, no deletion.

Movie Tools is a set of custom nodes, designed to simplify saving and loading batches of images with enhanced functionality like subfolder management and batch image handling.

Advanced image concatenation node for ComfyUI with flexible grid layouts, padding, and styling options.

With this node you can upload text files to input folder from your local computer.

Visual 3D camera-angle selector node with interactive hemisphere of 72 camera positions and distance controls.

AMD ROCm bf16 diagnostic tools for ComfyUI. Fixes black image issues caused by numpy's lack of bfloat16 support. Includes FP32 VAE decode/encode and debug nodes for latent, conditioning, and model inspection. Useful for AMD Strix Halo, older NVIDIA cards, Apple Silicon, and unified memory systems.

A custom ComfyUI node that unleashes Intel CPU + iGPU power via OpenVINO.

This repository contains a user-defined node for ComfyUI, currently there are nodes for capturing captions. But will be expanded in the future.

An extension for ComyUI that allows resetting a node's inputs to their default values

An extension for ComyUI to allow resetting a node's inputs to their default values. NOTE:Right click any node and choose 'Reset' from the context menu.

ComfyUI custom node that resolves a Qwen Image 2.1 Prompt Enhancer JSON hint (rewritten prompt + aspect ratio/size) into a ready-to-use prompt and exact width/height.

Introduces a node that tries to approximate the entire video using it's first frame (that we stylize) by warping it using optical flow extracted from the video. First we do image-to-image using the reference video's first frame and a depth controlnet. The generated object doesn't have to closely resemble the reference like on the demo. Then the generated image and the video frames are fed into the node and it returns a warped video. No ai models are used by the node. The example workflow uses SDXL but you can get it to work with any arch if you manage i2i.

Using rgthree's fast_group_muter and bookmark nodes, introduce the pyautogui library to simulate clicks and hotkeys, and run groups in sequence. screen manipulation is involved

A ComfyUI custom node that implements CoMPaSS for FLUX.1-dev models. CoMPaSS enhances the spatial understanding capabilities of text-to-image diffusion models.

Visual enhancement extension that adds animated arrows to links on hover and highlights Set/Get node pairs.

ComfyUI node for running evaluation after collecting latents with the EvalSampler, comparing against baseline BF16 latents. (Description by CC)

Another comfy implementation for the short video generation project PKU-YuanGroup/Open-Sora-Plan, supporting latest 1.3.0 and 1.2.0 and image to video feature, etc.

A comprehensive ComfyUI extension that enables text, image, video, and speech generation using any OpenAI-compatible API endpoint with custom URLs.

In-node Gaussian splat viewport (Spark) for ComfyUI with WYSIWYG image capture.

ComfyUI node that applies vintage/retro effects through JPG compression, color grading, film grain, vignette, and more to emulate more realistic photo aesthetics.

Advanced nodes for interaction with Olama (text, vision, image editing), with dynamic management of prompts via CSV.

Gradual IC-LoRA time-curve nodes for LTX-2.3 (ComfyUI). Modulate IC-LoRA strength over the output clip via a user-drawn curve, applied as a per-token runtime adapter.

Nodes:Multi-ControlNet Converter, List of Images, Convert PIL to Tensor (NHWC), Convert Tensor (NHWC) to (NCHW), Convert Tensor (NHWC) to PIL

ComfyUI sidebar extension for browsing, searching, and managing generated images and input assets. Scans input, output, and custom directories; indexes PNG/WebP metadata into SQLite for fast filtering by model, LoRA, sampler, prompt, and more; supports favorites, NSFW filtering, workflow loading, and an integrated fullscreen image viewer.

Modifed Sampler nodes that allow you to pause/resume generations. For those with older hardware running in CPU mode or old GPUs. Useful for managing thermals.

A ComfyUI custom node developed based on OmniAvatar, capable of generating video sequences with synchronized lip movements and facial expressions by inputting a portrait image, audio, and text prompt. The node parameters and invocation method are fully consistent with the official OmniAvatar inference.

NODES: SendToTostAI

A ComfyUI custom node pack that resizes images to an exact target canvas size with full control over scaling, cropping, and placement.

Implement circular padding VAE in ComfyUI from DiT 360 Paper

This is a ComfyUI custom node that extracts the last frame (image) from an input image batch. It is particularly useful when you want to obtain the final scene or frame in a video generation workflow.

ComfyUI Spectral is a ComfyUI custom nodes library based on the spectral, mainly used for visual processing of spectral files

A simple and effective ComfyUI custom node for filtering inappropriate text content, automatically detecting and replacing prohibited words while preserving the original format.

SeaCache for native ComfyUI FLUX, HunyuanVideo, Wan2.1, and MiniMax H3 models

ComfyUI's non-persistent (in-memory) image preview feature

🤖 ControlNet-aux-style preprocessor nodes for ComfyUI — Depth/Normal/Soft-Edge/Line/Pose map estimation, ported as real architecture code (not a thin wrapper) from Fannovel16/comfyui_controlnet_aux.

This custom node package for ComfyUI is designed to streamline your workflow with powerful file-counting capabilities.

Provide a prompt and get google image search results

Create a painting by colors image from an image

Run Clark Air Sana 1.6B (ternary ~1.58-bit, GemLite INT2 kernels) natively under ComfyUI's KSampler.

An extension for ComfyUI that adds utility functions and nodes not available in the default setup.

Load PSD / save PSD nodes with an interactive layer panel: load PSD or PSB files layer by layer (composite + per-layer image/mask outputs, all Photoshop blend modes), and save ByteDance Seedream layer PSD files with embedded smart-object layers. No Photoshop required.

These are custom nodes for ComfyUI for the loading and saving of metadata in XMP format. XMP metadata is embedded in the images created by these nodes; it travels along wherever the image does. Both macOS and Windows index XMP metadata automatically, making it searchable from the Finder on the Mac or the File Explorer in Windows. Apps like Photoshop or Lightroom (and presumably many others) expose XMP metadata and allow it to be edited.

A custom node for ComfyUI that integrates the Topaz Labs API for AI-powered image upscaling and enhancement.

ComfyUI custom nodes for Qwen text and vision-language models. Supports Qwen2.5/Qwen3, 4-/8-bit quantization, deterministic seeding, auto model download, and shared caching to improve workflow performance.

Smart resize, batch resize, resolution multiplier and prompt controller nodes for ComfyUI

Drag-and-drop workflow organization for the ComfyUI sidebar

A custom node extension for ComfyUI that integrates Google's Veo 3 text-to-video generation capabilities.

ComfyUI custom node: batch prompt queue from multiline/text file with incremental per-run consumption and {p} templating.

Anima-friendly first-pass resolution picker for ComfyUI. Outputs resolution, width, and height integers for image generation workflows.

A preset-driven Krea 2 Turbo sampler and 16-channel empty latent node for ComfyUI, with direct manual tuning and NegPiP-friendly model wiring.

Load and run GLM-OCR (zai-org/GLM-OCR) locally in ComfyUI. Supports text recognition, formula recognition, table extraction, and general image description. Picks up models from any ComfyUI checkpoint directory.

The primary goal of these nodes is to provide a way to access files generated by ComfyUI workflows, local to the machine running ComfyUI. These nodes should always return an updated list of files when triggered.

ComfyUI custom node for saving, searching, and reusing prompts across sessions.

Enhanced KSampler for ComfyUI with SmartResCalc noise passthrough and DazzleCommand integration. Based on RES4LYF's ClownsharKSampler.

Enhanced Preview Bridge node for ComfyUI with LayerCache architecture, dual-layer mask editing, and instant preview refresh. Extends Impact Pack's Preview Bridge with optional MASK input, mode selection, and WYSIWYG preview display.

ascii art preprocessors in ComfyUI

Prompt Style Selector with Preview, Multiple Selecting, Search and Local Base.

A set of lightweight utility nodes for ComfyUI that were originally crafted to support XYZ‑plot pipelines but are broadly useful in any workflow.

A collection of ComfyUI utility nodes. The Workflow Config nodes (WAN2.2 and LTX2.3) let you store multiple named presets — models, LoRA stacks, prompts, dimensions, and sampling parameters — and switch between them from a single dropdown, making it easy to test variations, manage scene setups, and iterate across versions without rebuilding your workflow or creating and maintaining multiple copies of it. Additional nodes: Lora Inspector (scans and displays safetensors metadata for any LoRA in your library), Check Null (detects null, None, NaN, or empty values), Null Audio Checker (detects silent/missing audio tracks from video inputs), and Abs Int (absolute value for integers).

This repository simply caches the CLIP embeddings and subtly accelerates the inference process by bypassing unnecessary computations.

Audio to midi functionality within ComfyUI

AMD/HIP native Triton W4A4 ConvRot nodes for ComfyUI.

Topaz Video AI nodes for ComfyUI. All processing runs locally.

Nodes:File Padding, Image Info, VAE Loader With Name

A custom node that saves images while displaying the seed value used in generation

A ComfyUI custom node package for generating professional interior design presentations using AI. This package integrates with Gemini/Ollama/OpenAI API to analyze floor plans and automatically generate PowerPoint presentations and PDF with AI-generated content and images. (Currently Support Chinese Only)

ComfyUI FlexPainter Wrapper

Custom ComfyUI nodes integrating Ollama to generate and enhance positive/negative prompts for Stable Diffusion workflows.

Real-time streaming for ComfyUI using the Livepeer gateway trickle protocol. Frames are published directly to orchestrators and outputs are pulled via trickle subscribe.

Sweep ComfyUI parameters and combine labeled frame batches into one comparison video.

NVIDIA ModelOpt integration for ComfyUI - Quantize and optimize diffusion models with INT8/FP8/INT4

A self-contained LoRA Slider node for ComfyUI designed to work alongside ComfyUI-Lora-Manager with bipolar strength sliders and batch LoRA loading.

One click, one whole coherent scene: the place, what is standing in it, and what it is doing - written as prose a text encoder can follow, where every detail agrees with every other one. A multi-genre scene engine: sci-fi (alien creatures, spacefarers, mechs, starships, stations, wrecks, worlds) fantasy (dragons, mythic beasts, giants, goblins, centaurs, elves, undead, golems, castles, airships) and horror (ghouls, vampires, banshees, cryptids, eldritch horrors, cultists, cursed objects, haunted houses, with a gore filter that starts on No gore), each drawn across a full morphology, with entities that wire across genres. Every field independently lockable.

Minimal ComfyUI custom node for per-model SageAttention opt-out and attention overrides.

High-quality pose detection plus interactive multi-layer image, depth, and pose compositing for ComfyUI. Uses RTMW-x / DWPose via rtmlib ONNX, and adds canvas-based image, depth, and text-layer composers with blend modes, drop shadows, and per-layer transforms.

ComfyUI nodes for UniPic3 multi-image composition and HOI

High-fidelity ComfyUI nodes for ERNIE-Image (SFT) and ERNIE-Image-Turbo using the native diffusers pipeline, with all critical text encoder and attention weight patches applied.

Custom ComfyUI nodes for extending video scenes with synchronized audio generation using LTXAVModel.

Generate, Browse, Organize, Share and Re-Use ControlNet Maps