ComfyUI Custom Extensions

Jacques' personal ComfyUI nodes (Gemini outpaint, padding, extras)

A custom node for ComfyUI that integrates Qwen 2.1 prompt rewriting for t2i and image editing

ComfyUI nodes for SoulX-Singer, a high-quality singing voice synthesis system.

Python script node for ComfyUI supporting math, calc, logic, and text editing with multi output. Handle conditional workflows, auto-select resolution, parse LLM outputs, and generate filenames — all in one flexible node.

Custom node for ComfyUI that performs color matching between reference and generated images with support for masking.

Shima - ComfyUI Workflow Creation and Sharing System

This node package contains automatic sampler setting according to model name in ComfyUI, adjusting image size according to specific constraints and some other nodes.

Custom nodes for convenient filtering image or string lists in ComfyUI workflow.

Nodes: CNCC ControlNet, CNCC Apply ControlNet. Saves comfyui_controlnet_aux preprocessors as reusable presets, and lets a single node run the preprocessor and apply the result as a ControlNet (a drop-in alternative to the native Apply ControlNet node). Includes a settings modal for creating/managing presets from a thumbnail gallery.

Load PSD files, rig layers with interactive R / MR / SW control points, and composite as IMAGE + MASK. Supports setup/pose modes, model/pose library, background options, and i18n (ja/en/zh).

Load and save JPEG XL images in ComfyUI with lossy/lossless compression and optional workflow metadata compression.

A clean, project-based prompt manager for ComfyUI that lets you store/load positive/negative prompts and typed variables per project.

Unofficial Implementation of Diff2Flow Method for ComfyUI.

15 productivity nodes: Kraken Unbound Prompt (vision-enabled prompt builder with Qwen2-VL), WAN Prompt Splitter (cinematic styling), Ollama Prompt Chat (LLM enhancement), LoRA Loader with CivitAI trigger fetching, Dual CLIP Loader (Flux/SD3/SDXL), smart KSampler (AMP handling for WAN/Flow/FP8), Empty Latent with aspect presets, Upscale & Tile Calculator for Ultimate SD Upscale, Resolution Helper, Image Processor, WAN Helper, and more.

A small ComfyUI custom node that provides a lightweight bypass for grouped nodes.

creating a custom node to test how to use Blocknet controlnets in comfyUI using logic from Diffsynth Studio.

Tools for processing equirectangular images, supporting conversion from equirectangular format to cubemap.

Advanced conditioning node for Flux2 Klein Kontext image editing workflows. Encodes up to 3 reference images via the Flux2 VAE and attaches them as reference latents for Kontext attention. Text is encoded through the Qwen3-8B LLM-based text encoder with optional instruct prompt wrapping. Supports per-image strength and resolution control, VAE tiling for high-resolution encoding, and bakes the Flux2 guidance scale directly into the conditioning.

This node loads all images from the selected folder and displays them in a grid. The user can click any image to automatically execute the ComfyUI workflow. Additionally, it supports many RAW image formats as well as HEIC formats.

Automatically loads trigger words from a .txt file with the same name as the LoRA. Supports offline use and custom-trained models.

Same Function as original ComfyUI_PuLID_Flux_ll repository, with FaceNet Implementation that is safe for commercial usage.

Drop-in replacement for Load Image that outputs the original filename as a string. Wire it into Save Image to auto-name your outputs. Preserves the real filename even after mask-editor / clipspace painting.

Nodes for generating multiple character prompt variations in a single queue run. Includes Character Prompt Manager and Variation Buffer.

Lightweight ComfyUI node for mosaic / pixelation restoration, based on lada.

ComfyUI nodes for batch processing, loops, and iteration workflows

Touch-first node/extension manager for ComfyUI: check updates, update nodes and core, install from a GitHub URL, select versions, fuzzy search — mobile-friendly.

Big corner grab-handles for resizing ComfyUI nodes and groups on touch devices.

A lightweight multimodal LLM node for ComfyUI: Ollama and OpenAI-compatible APIs over plain HTTP. SDK-free, zero extra dependencies.

ComfyUI nodes for working with Photoshop mockup files - automatically project images onto PSD smart objects while preserving perspective, rotation, and scale transforms for seamless mockup rendering.

Nodes related to video chat workflows

ComfyUI nodes to use [a/ChatGen](https://github.com/chengyou-jia/ChatGen)

The latest DIT architecture-based image generation model from Zhipu that supports Chinese text generation.

The ComfyUI version of [a/Qwen3-Omini](https://github.com/QwenLM/Qwen3-Omni)

ComfyUI nodes to use [a/QwQ-32B](https://huggingface.co/Qwen/QwQ-32B)

Support for directly inputting Chinese characters in the CLIP text box. The system will automatically translate them into English and replace the punctuation marks as well.

ComfyUI-GG is a collection of ComfyUI nodes designed to enhance productivity in image processing workflows. This plugin provides a set of custom nodes that perform various image manipulations and metadata extractions to streamline your tasks.

any to base64, and base64 to any.

Nodes to generate a pecha-kucha presentation in ComfyUI

ComfyUI wrapper nodes for Lumina-DiMOO. Please see the README for model weight download instructions.

OpenAI-compatible Chat Completions node for ComfyUI with text and image input support.

Auto crop continuous black borders from images

Dynamic subtitle effect nodes for ComfyUI - CapCut-style animated subtitles, CPU rendering with zero VRAM

Browse, search and manage ComfyUI's generated images with Hanaikada inside ComfyUI.

Enforces Qwen's training resolution buckets for precise Latent space generation.

Two simple to install nodes to get Live Portrait working in ComfyUI without the need for a fancy GPU (Replicate account needed).

Feature extension package for ComfyUI-Custom-Scripts.

ComfyUI expansion of the 🍅Tomato Crypt

ComfyUI utility nodes by AIKnowledge2Go.

Store and load ComfyUI media (image / audio / video) in an S3-compatible bucket, passing keys instead of asset bytes — built for using ComfyUI as an API across distributed installs. Great for building applications off of ComfyUI. Configure your bucket connection — copy `.env.example` to `.env` and fill it in.

Nodes:ComfyUI_Lora_List_With_Url_Loader. Load loras from urls and auto fetch them on web if they are missing.

ComfyUI nodes for Agate, LogoLabs' 260M-parameter text-to-image model: a single-file loader, a sampler that outputs SD 1.x LATENTs (img2img, live previews), an all-in-one 256x256 generator and a plan viewer that shows the thinker's 16x16 layout plan evolving over the denoising steps.

Nodes for controlling workflow and reducing VRAM fingerprint.

Custom nodes to offload and rapatriate models from cpu.

Token-aware spatial Multi-LoRA composition and transparent supersampling for Krea 2 in ComfyUI.

The Basic Drawing Board is a hybrid creative tool designed for ComfyUI.

ComfyUI LC Audio_Video Nodes by lonecatone23 — Create/Save Video, AV pipe, Load Audio cut preview, EQ, volume, channel, separate, VAE Decode wrapper, Get Latent Size. Companion to ComfyUI_LC123_nodes; separate Registry name.

A text translation node for ComfyUI using a local SLM (Microsoft Phi-4)

Bypass ComfyUI's torch.library.custom_op schema-registration bug that causes Flash Attention to silently fall back to SDPA (schema_.has_value() INTERNAL ASSERT FAILED).

Non-invasive SageAttention compatibility guard for ComfyUI. Auto-detects GQA, non-standard head_dim, and complex attention masks, falling back to native PyTorch SDPA without modifying ComfyUI core or virtual environment packages.

A ComfyUI custom node for combining LTXV conditioning latents from different sources.

ComfyUI nodes for GGUF-based Qwen3/Qwen3-VL text encoding and VLM inference for Z-Image

Advanced image list animation toolkit for ComfyUI with batch processing, AI-powered frame interpolation, and multi-format video encoding.

A ComfyUI custom node implementation of EdgeTAM (On-Device Track Anything Model) for efficient, interactive video object tracking.

Nodes: TAESD VAE Decode

Automatically upload ComfyUI output files to Google Drive with optional encryption. Supports OAuth and service accounts, real-time monitoring, cloud setups (RunPod), and includes cross-platform decryption scripts.

Automatically upload ComfyUI output files to Dropbox with optional encryption. Supports real-time monitoring, cloud setups (RunPod), and includes cross-platform decryption scripts.

Adds login, admin-managed user registration, API token generation, IP filtering, and per-account asset isolation to ComfyUI.

ComfyUI custom nodes for HappyHorse 1.0 video generation, including text-to-video, image-to-video, reference-to-video, video editing, and preview nodes.

ComfyUI custom nodes for the MiniMax H3 Video V2 API, supporting text-to-video, first/last-frame generation, multimodal references, H3-Context-IR, 2K regeneration, task management, download, and preview.

ComfyUI custom nodes for Gemma 4 multimodal AI

🥭 Mango Random Nodes - A collection of random file nodes for ComfyUI

Aholo Spatial Gen for ComfyUI: text-to-3D and image-to-3D space generation. Get API Key at https://labs.aholo3d.cn/api-keys (China) or https://labs.aholo3d.com/api-keys (Global).

This is a custom node aiming to run CogView4 on diffusers while there is no official implementation on ComfyUI. NOTE: You will need a updated version of diffusers and I don't know if updating it my break other stuff, so I advise you to make in a new instance of ComfyUI

A custom node for ComfyUI to load and use the LFM2-350M model trained to work as a prompt enhancer for z-image turbo or any other long token model.

Custom ComfyUI nodes for integrating RunningHub's AI image and video generation services into workflows. (Description by CC)

A custom node for ComfyUI that allows resizing of conditioning tensors, particularly useful for fixing size mismatches between CLIP Text Encode and CLIP Vision Encode outputs in SD3 workflows.

Multi-loop video generation with sharp output and character consistency

NODES: 'Load One Image with Name, Title, Description', 'Load Multiple Images with Name, Directory, Title, Description'

This node is a registry version of Kijais https://github.com/kijai/ComfyUI-GIMM-VFI

MetaMan bridges the metadata gap between AI image generation services, enabling seamless workflow sharing and model dependency tracking across platforms.

Node:Image Fit Calculator

Custom ComfyUI node for audio source separation using Meta Demucs v4 technology to extract voices, drums, bass, and other instruments with high fidelity. (Description by CC)

ComfyUI custom nodes for GPT-Image (gpt-image-2 and OpenAI-compatible image models). Text-to-image via /images/generations and image editing with up to 8 reference images via /images/edits. User-supplied base_url and api_key, no preset gateway.

A set of photo effects for ComfyUI, for the moment, only Bleach Bypass effect is provided, but more to come!

Useful custom nodes for ComfyUI by Meti: Outputs the total execution time (Generation Time) of your Nodes or workflow , Keeps a history of your generated images (up to 10) and shows them in a live-updating grid , Saves images with custom path and outputs filename & full path or Just Preview

Nodes:BBOXPadding, BBOXResize, ResizeKeepRatio.

Nodes: Primitive BBOX, BBOX Padding, BBOX Resize, ImageResize KeepRatio.

ComfyUI custom nodes for local VLM prompt generation with GGUF and Hugging Face backends.

StyledCLIPTextEncode is a node that enables you to build your prompts(both postive and negative) based on the selected style. It provides up-to 77 styles currently and has been tested on SDXL and SD1.5 checkpoints. It's ported from project [a/Style Selector for SDXL 1.0](https://github.com/ahgsql/StyleSelectorXL), which is only availabe on WebUI.

A simple workflow timer node for ComfyUI.

Removes background using Transparent Background

This repository contains a custom node for ComfyUI, 'Nano banana', designed for advanced image editing using the Google Gemini API.

A ComfyUI custom node package that provides two nodes for iterating through prompts and templates. Perfect for batch processing, testing multiple prompts, or creating variations.

ComfyUI nodes for LikeShare image obfuscation and restoration using a Hilbert-curve pixel shuffle.

A custom ComfyUI node using Together AI's Vision models (paid/free) to generate detailed image descriptions. Features advanced parameters, flexible API key management, and customizable prompts.

This node splits its self-attention Q to focus on nearby samples.

This node can easily switch between text or prompts by saving presets.