ComfyUI Custom Extensions

A ComfyUI extension that enables seamless sharing of AI-generated images and videos directly to Discord.

A powerful ComfyUI custom node for API-based image editing with multiple provider support

a custom node for [a/FireRedTTS](https://github.com/FireRedTeam/FireRedTTS)

Krea 2, 2.42x faster than BF16 on an RTX 3090 (up to 2.90x), at a third the size. Native INT8/W4A4 quantization with an SVDQuant low-rank branch, no calibration dataset needed. Turbo and base.

Custom ComfyUI nodes for ByteDance Lance — a unified 3B multimodal model supporting image and video understanding, generation, and editing in a single framework.

Turn creative ideas into Music 3 caption + lyrics via native VLM generation

A custom LoRA loader node for ComfyUI with advanced block-weighting controls for both SDXL and FLUX models. Features presets for common use-cases like 'Character' and 'Style', and a 'Custom' mode for fine-grained control over individual model blocks.

Nodes:Apply Invisible Watermark, Extract Watermark. Adds up to 12 characters encoded into an image that can be extracted.

Nodes:Easy2DDeforum (Chan)

A small collection of nodes intended for use with Lodestone Rock's Chroma model, for ComfyUI.

A modern, powerful, and easy-to-use Prompt Manager extension for ComfyUI.

Example node communicating between ComfyUI Javascript and Python.

Intelligent video frame selection node powered by Qwen3-VL-235B vision model. Automatically analyzes and selects the best quality frames from videos for further creative work.

Given min/max resolution constraints, this automatically suggests optimal dimensions while preserving aspect ratio. Ideal for Image to Image (I2I) and Image to Video (I2V) workflows!

Adds a RAM cache monitoring and management sidebar. Shows you which models are in VRAM and System ram, gives you an option to pin them to system ram so they can't be removed from cache, or force remove them yourself. Also has a button to free all VRAM without clearing system RAM cache, and an API endpoint to do the same thing. Can be configured to make ComfyUI wait for other processes to yield VRAM rather than plowing ahed and OOMing.

An universal media loader for ComfyUI (Size, Image, Audio, Video)

ComfyUI custom nodes to interact with the Finegrain API.

Join video clips end-to-end — instant, lossless, zero VRAM. Editor-style transitions: cut, dissolve, fades, wipes, slides. Works with loaded files or clips generated in-workflow.

Utility nodes for ComfyUI: video saving (H.264/H.265), universal switches, LoRA application, and animated latent preview.

Run any mflux/MLX model in ComfyUI on Apple Silicon, with a capability registry that forwards only the parameters each model accepts.

A ComfyUI node pack containing nodes for basic programming logic.

Tired of OOM errors and tedious file management? The AIIA Node Suite delivers the ultimate workflow overhaul, headlined by the AIIA Media Browser: a high-performance file hub fully integrated into ComfyUI. Instantly preview images, videos, and audio, and experience ultimate smoothness thanks to smart caching and virtual scrolling, even with tens of thousands of files. It also features a powerful fullscreen viewer, rich tooltip previews, and flexible sorting. Alongside the browser, our Video Combine node fundamentally solves memory bottlenecks in long video generation. By streaming frames from disk, you can create videos of virtually unlimited length, making OOM errors a thing of the past. The suite also includes OOM-safe FLOAT video generation, advanced audio processing, and other utilities, providing a complete end-to-end solution from content creation to asset management.

a [a/Facefusion](https://github.com/facefusion/facefusion)'s wrapper for ComfyUI custom node.

A custom node for ComfyUI that integrates OneReward model for high-quality image inpainting, outpainting, and object removal.

Add more endpoints to make easy for utilizing ComfyUI API.

A collection of post processing nodes for ComfyUI, dds image post-processing adjustment capabilities to the ComfyUI.

ComfyUI Blender Wrapper lets you call Blender in headless mode from ComfyUI (or any Python environment), run reusable pipeline stages stored inside files, and pass configs in/out via JSON. It's meant for tech-art / asset-pipeline tasks like cleanup, decimation, unwrap, baking, rigging, etc.

These are ComfyUI nodes to assist in converting images to paintings and to assist the Inspyrenet Rembg node to totally remove, or replace with a color, the original background from images so that the background does not reappear in videos or in nodes that do not retain the alpha channel in rgba images.

INT4 weight-only quantization (W4A16) via torchao. Activations remain FP16 — zero quality loss. Full LoRA: standard + LoKr, multi-LoRA stack (≤8), SHA256 JSON cache (<1s hit), GPU pre-hook bake-in. Auto-detects device on startup and installs correct torchao build (Intel XPU / NVIDIA CUDA / AMD ROCm). Verified: Krea2 Turbo, Flux2 Klein 9B, Z-Image, Boogu, Qwen-Image. Includes model analysis CLI.

This is a repository for using LoRTnoC (LoRA with hint block of ControlNet) on ComfyUI. NOTE:Please place the model file in the same location as controlnet. (Is this too arbitrary?)

LiquidTime is a simple yet powerful frame interpolation node for ComfyUI. Just input your sequence and desired frame count - the node handles all complex calculations and generates smooth in-between frames for you. A must-have tool for AI animation and video creation that lets you shape time like liquid.

ComfyUI Ultimate SD Upscale nodes with FLS tile sampling and optional Anima LLLite tile repair.

ComfyUI nodes to use [a/gemma-3-27b-it](https://huggingface.co/google/gemma-3-27b-it)

ComfyUI for [a/InternVL](https://github.com/OpenGVLab/InternVL)

Nodes:TCL EbSynth, TCL Extract Frames (From File), TCL Extract Frames (From Video), TCL Combine Frames, TCL Save Video (From Frames)

NODES: String Formatter, String List

ComfyUI nodes to use RAVE attention as a temporal attention mechanism. This differs from other implementations in that it does not concatenate the images together, but within the UNet's Self-Attention mechanism performs the RAVE technique. By not altering the images/latents throughout the UNet, this method does not affect other temporal techniques, style mechanisms, or other UNet modifications. For example, it can be combined with AnimateDiff, ModelScope/ZeroScope, or FLATTEN.

Nodes: RawText, RawTextCLIPEncode, RawTextCombine, RawTextReplace, Extension: m957ymj75urz.colors

Essential Nodes Pack for Images Processing for ComfyUI, with a live embedded preview inside the node (no queue) for supported chains.

Nodes:TextualInversionTrainingSDXL, TextualInversionTraining

OpenImageIO plugin for ComfyUI

This extension provides image generation features based on Fooocus.

A comprehensive node for ComfyUI that implements an image reference mechanism for Flux2-based image generation models.

Bilibili IndexTTS2 ComfyUI Version - Auto-downloads models, simple to use.

Custom javascript extensions for better UX for ComfyUI. Double click on image to open. It's convenient for checking images.

Native ComfyUI v3 nodes for the pixel-space AsymFLUX.2-klein 9B model from LakonLab (Asymmetric Flow Models, arXiv 2605.12964).

NK2E — experimental, unofficial in-context image editing for Krea 2 (community project; not affiliated with Krea).

A collection of custom nodes for ComfyUI

2 nodes for muting or bypassing a node by node ID. They are widget linkable and promotable to a Subgraph node as a switch to mute or bypass a node in it's own Subgraph, or any other Subgraph, or it can be used as a stand alone node anywhere

A prompt-rewriter node for Qwen-Image-2.1: pick a photographic style (camera language + mood) and a PE-T2I rewriter turns a short prompt into a long English one, plus a negative prompt and aspect-ratio dimensions. Not a model loader.

Video nodes for comfyui-pytorch360convert

Temporally coherent video generation and stylization for ComfyUI using optical flow. Port of SD-CN-Animation to ComfyUI custom nodes.

This plugin is compatible with the ComfyUI audio nodes.

A ComfyUI custom node wrapper for Meta's MeshFlow pipeline, providing state-of-the-art artist-like mesh remeshing directly within ComfyUI.

Hardened video workflow nodes for ComfyUI with safer output, metadata, and path handling

An empty latent size picker with support for Qwen Image, SDXL, and Flux models.

ComfyUI custom nodes for Ideogram 4 Magic Prompt and Ideogram 4 resolution selection

Expressive text-to-speech with zero-shot voice cloning. Ships with preset voices covering accents, ages, and emotions. Runs on 8GB VRAM.

A ComfyUI extension helper for the bult-in Queue Manager that gives you more control over the job queue. You can pause/resume, change the run order, hold individual items, and manually save/load the entire queue to disk for backup that survives a restart/reboot. It also auto-saves/restores the current running queue in case of unintended crashes. This extension does not load any new dependencies.

This is a collection of tools that I use to make my life easier when developing ComfyUI applications. It is a collection of tools that I have created to help me with my development process. I have decided to share these tools with the community in the hopes that they will be useful to others as well. I use this tools to further develop features for [a/https://shibiko.ai](https://shibiko.ai)

ComfyUI node for Semantic-aware Guidance based on the [a/paper](https://arxiv.org/abs/2404.05384) 'Rethinking the Spatial Inconsistency in Classifier-Free Diffusion Guidance'

Custom Loader for scaled_fp8 models with hybrid precision layers.

A powerful LoRA manager for ComfyUI with preview, tagging, and multi-node distribution support.

Run retro games inside ComfyUI via EmulatorJS and capture their screens for image generation workflows.

Just a reroute node with a tiny twist.

the sampler introduced by [a/hallatore](https://github.com/AUTOMATIC1111/stable-diffusion-webui/discussions/8457) code extracted from [a/smZNodes](https://github.com/shiimizu/ComfyUI_smZNodes).[w/NOTE:ComfyUI-dpmpp_2m_alt-Sampler is renamed to StableDiffusion-dpmpp_2m_alt-Sampler. Please reinstall.]

lynx:Towards High-Fidelity Personalized Video Generation, you can use this node in comfyUI with origin pipeline

A fork of ComfyUI-See-through by @jtydhr88, adding a custom node with the option to skip the head detail inference stage for faster processing.

Typed passthrough nodes to reduce wire clutter in subgraphs. Includes a multi-type dynamic passthrough, and utility nodes for batch switching among other various quality-of-life improvements.

Noise, conditioning, and perturbation utility nodes for ComfyUI (procedural noise, mesh drag, latent channel stats preview).

ComfyUI workflow for VEnhancer Inference

Marigold V2 in ComfyUI: single-step depth, surface normals, and albedo from one image.

All-in-One Face Fix KSampler for ComfyUI with YOLO detection and SAM segmentation

A ComfyUI extension for Live2D Pet integration and enhanced LoRA management.

A ComfyUI custom node that loads and stacks regular LoRAs and DoRA LoRAs with Flux/Flux2 and OneTrainer compatibility fixes.

Nodes: String, Int, Float, Short String, CLIP Text Encode (With Variables), String Format, Short String Format. This extension introduces quality of life improvements by providing variable nodes and shared global variables.

ComfyUI-SkyworkUniPic is now available in ComfyUI, Skywork-UniPic is a unified autoregressive multimodal model with 1.5 billion parameters that natively integrates image understanding, text-to-image generation, and image editing capabilities within a single architecture.

Put unlimited character LoRAs into one image, each locked to its own hand-drawn box — plus reference-image identity lock, cross-region attention dampening, per-region text conditioning, trigger-name auto-activation, and a sparse-token engine that only pays for the tokens that matter.

Nodes: PhotoMakerLoaderV2,PhotoMakerEncodeV2

ComfyUI sidebar extension — brings back the queue panel with image previews

Detect and resolve models missing from the active ComfyUI workflow graph, with multi-source resolution (Civitai / Hugging Face) and checksum-verified downloads.

This is a custom node for ComfyUI. Features:Removes white areas in the input image by making them transparent based on brightness, Outputs in black and transparent, Outputs in gray and transparent. This is a simple node with the above functionalities implemented. It also supports sequential processing.

A barebones ComfyUI wrapper for [a/PixelOE](https://github.com/KohakuBlueleaf/PixelOE). I cannot promise any support, if there is someone who wants to make a proper node, please do.

For ComfyUI Prompt_Tools

A properly implemented lycoris loader for comfyui.

A ComfyUI chat node based on SemiUI.

Lightweight CPU-based Text-to-Speech nodes for ComfyUI. Fast, efficient TTS running at 6x real-time on CPU without GPU. Supports voice cloning and 8 built-in voices.

This node allows you to cache/caching/store and reuse resized images, ControlNet images, masks, and texts. It avoids repeating heavy operations by loading previously saved files — saving time, memory, and processing power in future executions.

The CLIPSeg node generates a binary mask for a given input image and text prompt.

Simple implementation of [a/ByteDance Sa2VA](https://github.com/bytedance/Sa2VA) nodes for ComfyUI.

Using image caption models to extract prompts in ComfyUI

Colorize ComfyUI nodes with defaults per node, node category or via regex filtering.

Image and video nodes with compact controls: crop/rotate/pad editors for images and video, Load Image + Pad, Image Resize, Color Match, Compare A/B, Save Image and Save Video, Load Video, Frame Interpolate, Mask Refine, LaMa Inpaint, Crop/Stitch Inpaint, a LoRA Loader stack, Resolution Master, Seed, Workflow Note, Run Timer and utility nodes. Published on the Comfy Registry as ausboss-nodes.

Custom ComfyUI nodes by axces2000 — Audio Loader, Resolution Master, and more

Custom ComfyUI utility nodes for workflow enhancements.

A set of nodes for interfacing with the FastSDCPU webserver.

A ComfyUI custom node for MiniMax H3 Image And Reference To Video.

ComfyUI custom node to convert latent to RGB

A novel weighting scheme for token vectors from CLIP. Allows a wider range of values for the weight. Inspired by Perp-Neg.