ComfyUI Custom Extensions
Browse 9,420 ComfyUI extensions and 73,215 custom nodes.
a comfyui custom node for [a/UniAnimate](https://github.com/ali-vilab/UniAnimate)
a custom node for [a/F5-TTS](https://github.com/SWivid/F5-TTS)
Collection of image processing utility nodes for ComfyUI including intelligent cropping, aspect ratio scaling, and crop box restoration capabilities. (Description by CC)
It provides the capability to generate CLIP from an image input, unlike unCLIP, which works in all models. (To use this extension, you need to download the required model file from **Install Models**)
ComfyUI custom nodes for MiniCPM
GF Remove Background 2.0
Unofficial implementation of [a/MiniCPM-V](https://github.com/OpenBMB/MiniCPM-V) for ComfyUI
Custom Nodes by the team at WeirdWonderfulAI.Art. Line Count, Join String, Dither Image, Image Batch Loader, Prompt Writer
Nodes:FaceSwapNode. Very basic custom node to enable face swapping in ComfyUI. (roop)
A collection of custom nodes for working with and converting between and working with 360 degree equirectangular images, cubemap, and perspective images. Also includes Circular padding Conv2d options for models and VAEs, along with nodes to fix seams and polar artifacts. Panoramic 360 images are also sometimes known as VR photography (virtual reality), HDRI environments (ex: skyboxes), image spheres, spherical images, 360 pano.
Image Adjustments node for ComfyUI with minimal requirements, uses PyTorch for image manipulation operations.
Send notifications when a workflow completes.
ComfyUI custom nodes for RVC related inference and image generation
An advanced custom node for ComfyUI that provides optimized access to Wan2.1, a state-of-the-art video foundation model suite. The WanVideoKsampler node features intelligent memory management to enable higher resolution outputs and longer video sequences, even on consumer-grade hardware.
Directly upscaling inside the latent space. Model was trained for SD1.5 and drawn content. Might add new architectures or update models at some point. This took heavy inspriration from [city96/SD-Latent-Upscaler](https://github.com/city96/SD-Latent-Upscaler) and [Ttl/ComfyUi_NNLatentUpscale](https://github.com/Ttl/ComfyUi_NNLatentUpscale).
A custom ComfyUI node that preserves High Dynamic Range (HDR) data during VAE decoding, bypassing the default 0-1 range clamping. It also includes a companion 'Linear EXR Export' node for creating professional-grade linear EXR files.
One-click MiniMax H3 long-video generation node: chunked inference, inter-chunk keyframe continuation anchoring, and timeline-based prompt slicing for seamless long videos on limited VRAM.
This node speeds up Flux-based models in ComfyUI by using INT4 (W4A4) quantization on RTX 30 and 40 series cards, delivering faster inference and lower VRAM usage.
Node to pause a workflow with a continue button
These nodes will be placed in comfyui/custom_nodes/aegisflow and contains the image passer (accepts an image as either wired or wirelessly, input and passes it through. Latent passer does the same for latents, and the Preprocessor chooser allows a passthrough image and 10 controlnets to be passed in AegisFlow Shima. The inputs on the Preprocessor chooser should not be renamed if you intend to accept image inputs wirelessly through UE nodes. It can be done, but the send node input regex for each controlnet preprocessor column must also be changed.
ComfyUI-Book-Tools is a set o new nodes for ComfyUI that allows users to easily add text overlays to images within their ComfyUI projects. This Node leverages Python Imaging Library (PIL) and PyTorch to dynamically render text on images, supporting a wide range of customization options including font size, alignment, color, and padding. Loop with any parameters (*), prompt batch schedule with prompt selector, end queue for automatic ending current queue.
Multi-GPU Ulysses sequence-parallel inference for MiniMax-H3 video and audio generation across 2/4/7/8 GPUs. Requires ComfyUI commit c194dd00 or newer and Linux or WSL2 with NCCL; native Windows is not supported.
Welcome to ComfyUI Prompt Preview, where you can visualize the styles from [sdxl_prompt_styler](https://github.com/twri/sdxl_prompt_styler).
ComfyUI Img2Img-Turbo
The official ComfyUI TextEncodeQwenImageEdit node simplifies prompt engineering. This node supports custom system prompts to enhance flexibility.
A node suite for downloading audio and video from youtube as we all sevral useful video utilits such as a final frame selector and a node that merges two videos into one.
Minimal single-image MiniMax H3 photo editing with native or semantic references
A powerful ComfyUI custom node for advanced face occlusion, segmentation, and masking, leveraging state-of-the-art face detection (insightface buffalo models) for robust and accurate results.
Connect compatible nodes without scrolling across your graph.
Run Segformer at lightning speed for image/video segmentation.
Prompt editing, ANIMA prompt correction, NAIA prompt integration, LoRA preset management, wildcard expansion, AiO generation, and ANIMA/Spectrum workflow helpers for ComfyUI.
This tool will turn entire workflows or parts of them into single integrated nodes. In a way, it is similar to the Node Templates functionality but hides the inner structure. This is useful if all you want is to reuse and quickly configure a bunch of nodes without caring how they are interconnected.
A Wrapper and a set of Custom Nodes for using RenderFormer as a 3d Environment in ComfyUI.
ComfyUI custom nodes for rednote-hilab dots.tts with model loading, generation, voice cloning, Whisper transcription, progress tracking, and AIMDO memory visualization
A ComfyUI production pack with Studio workflow cores, Outfit Forge, a portable Templates/Prompts/Outfits/Scenes Creative Library, a visual LoRA Library, and safe LoRA organization.
some samplers by spawner
A custom node for ComfyUI that allows you to create iPhone-compatible Live Photos from videos. This node can convert video sequences into Live Photo format, with the ability to select key frames and customize the output.
A simple and efficient ComfyUI integration for DreamID-V, a model for identity-preserving video generation that animates reference images using motion from pose videos.
Kani TTS. Generate natural, high‑quality speech from text
a comfyui custom node for [a/Retrieval-based-Voice-Conversion-WebUI](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI.git), you can Voice-Conversion in comfyui now! NOTE: make sure ffmpeg is worked in your commandline for Linux
ComfyUI implementation of FLUX.2-dev-Fun-Controlnet-Union from Alibaba's VideoX-Fun
ComfyUI 官方的 `TextEncodeQwenImageEdit` 节点简化了提示词工程,这个节点支持自定义系统提示词以增强灵活性
AI autoprompter for Ideogram 4's structured JSON caption format — local Qwen3-VL or Gemini build the caption, then edit it on a visual bbox canvas.
Inpainting with layers inside ComfyUI: one node holds the image, the layers, the selection and the prompt. The selected region goes out as a crop to any inpaint chain and the result comes back as a new layer. Selection tools (brush, lasso, polygon, magic wand, SAM2 objects, SAM3 text), reference layers for Flux.2 multi-reference editing, boxes in the prompt (FLUX 3's edit rows on a box_prompt output, core bounding-box elements for FLUX 3 Image and MOSS-Image), RMBG cutouts, film grain and LUT filter layers, prompt upsampling, PSD and ORA export. The same editor is also a standalone desktop app, Scumble (https://github.com/DenRakEiw/scumble), which renders through ComfyUI with this pack or through API providers.
Automatically benchmark and optimize attention in diffusion models. 1.5-2x speedup on RTX 4090, up to 4x on video models.
This custom node gives you total control over where your models run and where they live in memory. It solves common Out of Memory (OOM) errors, enables multi-GPU workflows, and fixes persistent crashes when trying to run modern workflows (like Flux/SD3) on CPUs or secondary GPUs.
This is a bunch of nodes for ComfyUI to help with sound work.
ComfyUI nodes wrapping the [a/deepface](https://github.com/serengil/deepface) library.
Improved qwen image editing accuracy
Long MiniMax H3 videos clip by clip with continuity: motion/identity transfer (Ref2VA) and style transfer / retexture (guide latent + StyleTransfer LoRA), resumable, one clip in VRAM at a time.
An experiment about combining multiple LoRAs with [a/DARE](https://arxiv.org/pdf/2311.03099.pdf)
A real LLM chat panel that lives inside ComfyUI — not a one-shot prompt box.
ComfyUI nodes for MiniMax H3: INT8 X2 VAE and asynchronous NVENC saving. 5-second 1088x1920 video with audio in about 60 s on an RTX 4070 12GB, about 30% faster than the FP16 X2 VAE.
ComfyUI custom nodes for OmniGen project. Low VRAM, preview, separated VAE and more
Nodes:Send Webp Image to Eagle. This is an extension node for ComfyUI that allows you to send generated images in webp format to Eagle. This extension node is a re-implementation of the Eagle linkage functions of the previous ComfyUI-send-Eagle node, focusing on the functions required for this node.
ComfyUI speech recognition plugin based on Qwen3-ASR with multi-language support and forced alignment for timestamps.
This is a ComfyUI extension that provides additional API endpoints functionality, primarily designed to support Comfy Portal - a modern iOS client application for ComfyUI.
Try [a/OmniParser](https://github.com/microsoft/OmniParser) in ComfyUI which a simple screen parsing tool towards pure vision based GUI agent.
Adjusts a latent image before the VAE decoding step in order to improve your image quality.
Adjusts a latent image before the VAE decoding step in order to improve your image quality.
This is a custom node for ComfyUI that can be used to generate videos from either a starting frame, an end frame or both; with the Wan2.2 5B model (which uses the new Wan2.2 VAE, unlike Wan 2.2 A14B model wich uses the old Wan2.1 VAE).
X-FluxAgent turns ComfyUI into a smart, AI-powered agent capable of building software, automating tasks, and even managing your daily workflows — all with natural language prompts, no coding experience needed.
Generate phase-preserving structured noise for the input image.
Image scoring nodes for ComfyUI using PickScore with a batch of images to predict which ones fit a given prompt the best.
A ComfyUI custom node that simply integrates the [a/animate-anyone-reproduction](https://github.com/bendanzzc/AnimateAnyone-reproduction) functionality.
A ComfyUI custom node that provides a LoRa loader with persistent trigger word storage. Automatically saves and loads trigger words for each LoRa model, making your workflow more efficient.
Nodes for using models from online providers. Flux, Auraflow, SoteDiffusion, etc.
Janky implementation of [a/DiffuseHigh](https://github.com/yhyun225/DiffuseHigh/) for ComfyUI. Enables generating at resolutions higher than what the model was trained for without requiring model patches.
Essential string manipulation nodes for ComfyUI.
A ComfyUI wrapper for [a/Kokoro-onnx](https://github.com/thewh1teagle/kokoro-onnx)
A collection of custom nodes for ComfyUI, featuring high-precision color matching (MKL, Wavelet, HM), latent size adjustment, and metadata-aware image saving.
ComfyUI node pack with latent color match and image adjust tools, a universal neural-network latent upscaler (SD1.5/SDXL/Flux/Wan2.2), Flux LayerDiffuse transparent image generation, a multi-image aspect ratio composer, and Flux 3 API nodes. Integrates the standalone Latent_Nodes and WAN_NN_Latent_Upscale packs (DRE-tagged to coexist without conflicts).
MeanCache: Training-free inference acceleration for Z-Image Flow Matching models
Node for the FASHN Virtual Try-On API. Requires an API Key from fashn.ai
The Artisan’s Bridge between Adobe Photoshop and ComfyUI. Delivers real-time bidirectional canvas sync, an Edge-Cloud hybrid engine (RunningHub & BizyAir), interactive 3D/VR previews, and an essential suite of custom nodes for lossless image alignment, pause-control, and workflow logic.
A node that allows you to switch between execution flow
This is a single node which can take the POSE_KEYPOINT output from the OpenPose extractor node, parse it, and return x,y,width,height bounding boxes around any elements of the OpenPose skeleton
Advanced camera control prompt generator for ComfyUI that reads camera information from 3D nodes and generates precise, professional camera control prompts.
Local LLM session nodes for ComfyUI using GGUF and llama.cpp, supporting Llama, Mistral, Qwen, DeepSeek, GLM, Gemma, Phi, LLaVA and gpt-oss, enabling both user–model chat and model-to-model dialogue without external runtimes like Ollama.
Run ComfyUI workflows on multiple local GPUs/networked machines with options to edit the json values within comfyui. Original repo: [a/city96/ComfyUI_NetDist](https://github.com/city96/ComfyUI_NetDist)
Channel curves custom node for ComfyUI. RGB, R, G and B, luma and saturation curves adjustments. Save and load presets. Real-time curves adjustment tool directly within the user interface. Precise, interactive control over the tonal range of both image channels and masks, using a GPU-accelerated PyTorch backend for instant feedback.
Registered outpainting nodes for the yijunwang2/krea2-outpaint LoRA on Krea 2 Turbo. Places the source reference into the target latent grid at a bbox (RoPE frame i+1) with isolated-ref KV cache, then restores exact pixels with a feathered seam. Reference-attention machinery adapted from ostris/ComfyUI-Krea2-Ostris-Edit (MIT); placement geometry from the outpaint release (Apache-2.0).
Comprehensive ComfyUI custom node suite featuring Smart Prompt with wildcards, a pipe ecosystem, universal type converters, image/video utilities, and workflow helpers. V3 API, Nodes 2.0 compatible.
Nodes:Detect By Label.
Nodes:SDXL Empty Latent Image. An extension node for ComfyUI that allows you to select a resolution from the pre-defined json files and output a Latent Image.
A video editor for MiniMax H3 inside a single ComfyUI node. Keyframes with independent strength dials (guide the motion instead of pinning it), waypoints the clip passes through, and a fullscreen timeline editor with filmstrip cards, framing tool, cast members that persist across clips, and free Creative-Commons image/audio search built in. Continue a clip with real continuity: the previous render's tail frames AND audio are pinned at the new head, so motion and the actual waveform carry through the join — with an auto mode that chains clips hands-free. Finished clips land on a reel you can trim, crossfade, level, score with a soundtrack lane plus three fx lanes, and export as one video. Also includes v2v restyling, soft denoise zones (differential diffusion with per-frame masks), regional prompting and a temporal LoRA blend.
A drop-in replacement for ComfyUI's stock ReferenceLatent node, with per-image strength, per-image timestep gating, per-image MediaPipe auto-masks, megapixel cap on attention cost, and 1–4 image inputs in a single node.
Video transition effects for ComfyUI using WebGL rendering with 100+ transition variants across categories like dissolve, wipe, zoom, blur, glitch, and more.
[a/AniCrafter](https://github.com/MyNiuuu/AniCrafter): Customizing Realistic Human-Centric Animation via Avatar-Background Conditioning in Video Diffusion Models, you can try this methods when use ComfyUI.
Nodes to run native weights of LTX's OmnimatteZero. Workflow: https://github.com/tpc2233/ComfyUI-TP-OmnimatteZero/tree/main/workflows
Nodes for Google Gemini API, focusing on backward compatibility and stability within ComfyUI.
Video clip mixing
This project is an unofficial ComfyUI implementation of [a/Thera](https://github.com/prs-eth/thera) (Aliasing-Free Arbitrary-Scale Super-Resolution with Neural Heat Fields)
Custom Nodes for ComfyUI to Projects Texture into your 3D models.
a custom node for [a/FancyVideo](https://github.com/360CVGroup/FancyVideo)
NODES: AGSoft Empty Latent, AGSoft ImageRes, AGSoft ImageResMP, AGSoft ImagegPad for Outpainting, AGSoft ImagePad for OutpaintingAdv, AGSoft Image Crop, AGSoft Image Concatenate, AGSoft Image ConcatenateFromBatch, ...
带时间戳、标点符号,自动语音识别。给视频自动添加字幕。
Automated Speech Recognition for Chinese.
ComfyUI KSampler Matrix Lab is a custom node for comparing multiple sampler and scheduler combinations in one labeled image grid.
Architecture-aware LoRA loader for Z-Image Turbo (Lumina2) in ComfyUI. Fixes silent key mismatches by auto-fusing separate Q/K/V into Z-Image's fused QKV format and remapping output projections.