ComfyUI Custom Extensions

Timeline-based director node for MiniMax H3 — visual CUT/segment editor, multi-character reference images, reference video + reference audio (with per-clip audio pairing for lipsync/voice reference), First/Last Frame mode, automatic chunking with a hybrid hard-lock continuation option, and a Seed Hunt toggle that scouts 4 candidate seeds in one run for use with Muse Minimax Refine's second-pass upscale.

Nodes: Format String, Join String List, Load Preset, Load Preset (Advanced), Const String, Const String (multi line). Add useful nodes related to prompt.

This repository is a ComfyUI port of the WithAnyone model introduced in the paper. [a/WithAnyone: Towards Controllable and ID-Consistent Image Generation (2025)](https://arxiv.org/abs/2510.14975) Original implementation: [a/Doby-Xu/WithAnyone](https://github.com/Doby-Xu/WithAnyone).

Audio/Video handle management for ComfyUI workflows Add and remove stabilization frames with synchronized audio for video diffusion models.

An unofficial ComfyUI custom node integration for High-quality Text-to-Speech and Voice Conversion nodes for ComfyUI using ResembleAI's ChatterboxTTS with unlimited text length!!!.

A user-friendly way to selectively transform parts of your starting image in WAN Image-to-Video generation with advanced masking and person detection.

Frequency-Decoupled Guidance (FDG) for ComfyUI and ComfyUI-LTXVideo that improves upon standard Classifier-Free Guidance by applying separate guidance scales to different frequency components.

Nodes: tri3d-extract-hand, tri3d-fuzzification, tri3d-position-hands, tri3d-atr-parse.

Custom nodes that let you quickly load multiple images via a multi-select dialog with preview. The images are instantly uploaded to the input folder and can be output either as a list or a batch.

Nodes: Yolov8Detection, Yolov8Segmentation. Deadly simple yolov8 comfyui plugin

a comfyui custom node for [a/DH_live](https://github.com/kleinlee/DH_live)

Custom nodes for loops, flow control, lazy switches, image processing, and latent manipulation.

Custom nodes for for loops, while loops, lazy nodes and switches etc.

This custom ComfyUI node, Find Perfect Resolution, calculates an optimal output resolution for an input image while preserving its aspect ratio and ensuring dimensions are divisible by a specified value. It is designed to work seamlessly in ComfyUI workflows, particularly for resizing images with nodes like 'Resize Image v2'.

FunASR wrapper for ComfyUI

Per-block LoRA weighting for Flux: scan all 57 transformer blocks (D00-D18, S00-S37), batch-sweep strengths, and render labeled comparison grids.

Per-block LoRA weighting for DiT models (FLUX.1, Qwen-Image, SD3.5 Large): scan every transformer block, sweep strengths, and render labeled comparison grids.

This node displays the positive and negative prompts of a loaded ComfyUi image.

Adds an Image Loader node that also shows images in subfolders of the default input directory

ComfyUI-Free-GPU provides a node for releasing RAM and VRAM in ComfyUI.

Custom nodes for MiniCPM language models in ComfyUI. Provides advanced text generation and image understanding functions.

A simple but powerful node for ComfyUI that displays text input in a readable format. Perfect for viewing outputs from text generation nodes, prompt builders, interrogators, and more.

A comprehensive Vision-Language Model (VLM) integration system for ComfyUI with more intelligent prompt optimization, object detection, template support, and performance optimizations. Optimized for Wan2.1, Flux Kontext, and general purpose. Goes well with my other project, an MLX/llama.cpp server with hot swappable models and ollama api compatibility, (heylookitsanllm)[a/https://github.com/fblissjr/heylookitsanllm](https://github.com/fblissjr/heylookitsanllm)

This extension provides various nodes to support multimodal workflows.

FL DiffVSR - Diffusion-based video super-resolution nodes for ComfyUI. Features 4x upscaling with temporal coherence using Stream-DiffVSR for smooth, artifact-free video enhancement. Supports text-guided upscaling, chunked processing for memory efficiency, and automatic model downloading from HuggingFace.

The ultimate suite of streamlined ComfyUI nodes for creative professionals.

Collection of scripts, patches, and custom nodes for ComfyUI

A set of custom nodes and aiohttp server for a simple ComfyUI browser and mobile device experience for any t2i image workflow.

Tiled upscaling for ComfyUI that reads the whole image first, then writes a separate accurate prompt for every tile.

Custom node plugin for ComfyUI to experiment with Truncated Jump Sampling (TJS) for optimized diffusion sampling. (Description by CC)

Talking avatars Heads for the IF_AI tools integrates dreamtalk in ComfyUI

Talking avatars Heads for the IF_AI tools integrates dreamtalk in ComfyUI

ComfyUI-Easyai is a powerful extension for ComfyUI that enables users to share workflows and models to easyai.

ComfyUI-Transformers is a cutting-edge project combining the power of computer vision and natural language processing to create intuitive and user-friendly interfaces. Our goal is to make technology more accessible and engaging.

Temporary repository for testing candidate for native ComfyUI loop nodes. (Description by CC)

Encapsulate the comfyui workflow as a gradio webui

Nodes: Remove Image Background (abg). A Anime Background Remover node for comfyui, based on this hf space, works same as AGB extention in automatic1111.

ComfyUI Custom Node for 'Golden Noise for Diffusion Models: A Learning Framework'. This node refines the initial latent noise in the diffusion process, enhancing both image quality and semantic coherence.

A MiniMax H3 guide, local filmstrip, and endpoint-anchor timeline for ComfyUI.

Model-aware semantic face repair for single-face and multi-face ComfyUI workflows.

A set of custom nodes for frame interpolation and video processing in ComfyUI.

ByteDance Bernini-R (Wan2.2-T2V-A14B + source-id RoPE + APG multi-condition guidance) for ComfyUI: t2v/t2i, image & video editing, reference-to-video.

AMD gfx12 (RX 9070/9070 XT) memory-efficient Sage Attention for MiniMax H3. Uses native HIP kernel, saves 33% VRAM. Mirrors KJNodes NVIDIA patch.

ComfyUI implementation of [a/FlashFace: Human Image Personalization with High-fidelity Identity Preservation](https://github.com/ali-vilab/FlashFace) NOTE: You need to downalod models manually.

ComfyUI nodes for FlashPortrait that enable infinite-length portrait animation generation driven by video with automatic model downloading and flexible input options.

The Comfyui-PT-Keyframe-Camera is a powerful tool designed to streamline the process of creating animations with keyframe cameras. It integrates seamlessly with existing workflows, allowing you to focus on your creativity without getting bogged down by technical details. This tool is perfect for animators, filmmakers, and anyone looking to add depth to their projects.

Default 16GB VRAM UNO in context generation ComfyUI-node, using RED-UNO FT model

Control Freak gives you physical control over your ComfyUI workflows by mapping MIDI controllers, gamepads, and other input devices to any node parameter or UI element. Transform your image generation experience with tactile, responsive control. Ever heard of 'flow state'?

This node integrates the face-swapping capabilities from Deep Live Cam into ComfyUI, allowing you to perform real-time face swapping on images and video streams.

This code gives you the quick access to 14 million text-to-image prompts from the DiffusionDB dataset.

You can call Chatglm's API in comfyUI to translate and describe pictures, and the API similar to OpenAI.

ComfyUI Video Anti-Aliasing Pack

24GB-optimized VDN-H3 hybrid attention for MiniMax-H3 in ComfyUI

A media viewer node for ComfyUI. Auto-captures all generated media into a grid for easy viewing, sorting, and comparing.

Nodes:MuseVImg2Vid (comfyui_musev_evolved) NOTE: Download [a/MuseV](https://huggingface.co/TMElyralab/MuseV) to ComfyUI/models/diffusers

A simple node to load image from local path or http url. You can find this node from 'image' category.

This release introduces Vantage Nodes, a powerful utility and infrastructure node pack for ComfyUI, designed for advanced workflows, batch processing, and experimental pipelines.

Nodes: WAS_PFN_Latent. Perlin Power Fractal Noisey Latents

A video transition effects plugin implemented using a Chromium headless browser and Python.

A collection of custom nodes for ComfyUI focused on enhanced sampling, model optimization, and quality improvements.

the comfyui custom node of [a/MuseTalk](https://github.com/TMElyralab/MuseTalk) to make audio driven videos!

Node Pack: PromptChecker for token toggling, KoboldCPP API, ModelMerging, Telegram-Bot-API, and more

A node that assists in one click generation of prompts (for image and video generation, etc.) in Comfyui.

A node in comfyui for one-click assisted prompt generation (for image and video generation, etc.).

ComfyUI-Krea2-Projector-Tuner is a specialized custom node package for ComfyUI that allows you to patch and tune the layers of Krea2 models in-memory.

Resource monitor for Nvidia within ComfyUI

ComfyUI custom nodes to apply various image processing techniques.

This package provides three custom nodes designed to streamline workflows involving API requests, dynamic text manipulation based on API responses, and image posting to APIs. These nodes are particularly useful for automating interactions with APIs, enhancing text-based workflows with dynamic data, and facilitating image uploads.

ComfyUI Champ

Smart Face Alignment and Pasting Node

Advanced voice cloning with 80+ emotion presets and multi-emotion mixing for ComfyUI. (Description by CC)

Krita-style inpainting inside ComfyUI: a layered canvas with selection tools (brush, lasso, polygon, SAM2 objects, SAM3 text), the selected region goes to any inpaint chain and comes back as a layer. Reference layers for Flux.2 multi-reference editing, RMBG cutouts, film grain and LUT filter layers, prompt upsampling.

ComfyUI wrapper for [a/Pixal3D](https://github.com/TencentARC/Pixal3D) — Tencent's SIGGRAPH 2026 single-image to PBR-textured-3D pipeline. Windows + NVIDIA RTX 30/40/50 (≥ 16 GB VRAM). One image → textured PBR mesh in ~3-5 min. Built on top of [a/ComfyUI-Trellis2](https://github.com/visualbruno/ComfyUI-Trellis2). Note: Pixal3D itself is academic / non-commercial only and NOT for EU use — see NOTICE.md.

A 70+ field character randomizer node pack for ComfyUI with a constraint engine, themed archetype presets with varying costumes, and dual prose/JSON output.

Contains ImageSBSConverter node to convert an image into a side-by-side image.

Ollama-driven routing nodes and curated workflows for ComfyUI with checkpoint, LoRA, and Flux split-component flows plus image-guided variants using vision analysis.

ComfyPanel — a ComfyUI and Photoshop plugin (including custom ComfyUI nodes) that brings local and cloud AI workflows into your creative workspace. Supports RunningHub and BizyAir, offers bidirectional sync between Photoshop and ComfyUI, and automatic recovery of asynchronous cloud results.

This is a custom node pack designed for ComfyUI that seamlessly integrates the powerful LM Studio into your workflow, enabling you to perform a variety of tasks such as text generation and image understanding (Vision) using locally run LLMs.

This is a ComfyUI custom node implementation for image editing using the Step-1 model architecture, specifically adapted for reference-based image editing guided by text prompts.

ComfyUI native implementation of [a/AI-Assistant](https://github.com/tori29umai0123/AI-Assistant).

Nodes to use [a/faster-whisper](https://github.com/SYSTRAN/faster-whisper) for transcription.

ComfyUI LC123 Nodes by lonecatone23 — Image Label, Skin Upscale, Bypass Relay, Image/Mask Pass, Widget To String, Save Image + Metadata, Any Empty Bool/Int/Float, Int Split, LC Batch Image, Tone Match, Color Match mask, MiniMax H3 pipe, Image-Mask Resize, LC Lighting Control, Photo Style, Skin Beauty, Sharpen Pro, prompt builder, sampler configure, Sigma Curve, Sigma Resample, image FX, pipes, notify.

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.

Mask cutout based on Segment Anything.

The first PuLID (face identity preservation) implementation with full Chroma model support for ComfyUI. Automatically detects and adapts to both FLUX and Chroma models. Fork of PuLID-Flux-Enhanced with comprehensive Chroma compatibility fixes.

A comprehensive node pack for ComfyUI that provides powerful image effects, pattern generation, and animated pattern creation capabilities.

The 'Eses Image Lens Effects' node is a multipurpose node for ComfyUI designed to simulate a variety of lens characteristics. It combines several typical effects into a single, convenient node, allowing to add realistic or stylistic lens distortion, chromatic aberration, post-process scaling, and a highly configurable vignette.

ComfyUI nodes for game asset preparation workflows

ComfyUI-native Bernini in-context conditioning for Wan 2.2 with external LLM prompt-enhancement helpers.

[a/Onediff](https://github.com/siliconflow/onediff) ComfyUI Nodes.

Use DICE-Talk in ComfyUI，which is a method about Correlation-Aware Emotional Talking Portrait Generation.

This node allows you to use ID_Animator, the zero shot video generation model

Making Avatars InteractTowards Text-Driven Human-Object Interaction for Controllable Talking Avatars

Nodes: ComfyUI Mexx Styler, ComfyUI Mexx Styler Advanced

A node for ComfyUI that takes an input image and clips the color channels independently to increase contrast and alter color cast. This is a reinterpretation of PhotoShop's 'Auto Tone' algorithm.

A suite of custom ComfyUI nodes built to evaluate and diagnose the technical qualities of images—especially those generated by AI models. Rather than creating visuals, these tools measure them, offering precise insights into sharpness, noise, exposure, color balance, and more.

Folder-backed prompt library and multi-track scheduler for ComfyUI, plus optional NovelAI (NAI) nodes. Browse/edit wildcards and JSON cards with LoRAs, sparse generation settings, and sparse nai fields; auto-apply models/params and embed wildcard LoRAs per queue job; NAI serial batch with resume; reuse image metadata; save with A1111-style parameters.

High-Frequency Detail Sampling based on Res Sampling for ComfyUI

ComfyUI custom node，use Deepseek v3 to classify the input tags