Image Generation & Editing

Make MiniMax H3 faster in ComfyUI on one 24 GB card: a lossless conditioning + latent disk cache (encode once, sample many seeds), plus measured SageAttention and EasyCache tradeoffs — including EasyCache's hidden quality cost and the end_percent fix.

ComfyUI custom node: build a character portrait prompt from structured fields (Character Studio).

A GPU/CPU-aware model proxy daemon designed to manage multiple LLMs and other AI tools (like ComfyUI) efficiently.

macOS natural-language client and on-demand RunPod ComfyUI worker for Mira Studio

ComfyUI support for DepthAnything V3 model

How Python does AI. Agents, realtime voice, image generation, embeddings. Every model, every interface, typed end to end.

ComfyUI node that automatically converts Chinese prompts and image/video references into the official standard format for MiniMax-H3 video generation.

Qwen's most powerful open-source image generation model

ComfyUI/Blueprints-style node graph editor component for Streamlit — typed ports, drag-to-connect

an modified version of comfystudio that works with comfyui from diferent sources and works from a remote location.

ComfyUI custom nodes for spatially upscaling a MiniMax H3 video latent between two sampling passes (low-res -> upscale -> high-res)

MiniMax-H3 NF4 text-to-video with synchronous audio as ComfyUI nodes. Local inference, no API key, 70% disk savings over BF16.

ComfyUI custom node that reproduces Automatic1111-style subseed (variation seed) by interpolating noise tensors from two seeds.

SGLang is a high-performance serving framework for large language models and multimodal models.

Fizgig — LoRA & Fine-tune Studio for Klein 9B, Krea 2, MiniMax H3 & Qwen Image 2.1: train, fine-tune, profile, repair and extract LoRAs & LoKRs

All-in-one KREA 2 Film Studio custom node for ComfyUI with T2I, I2I, directed control, LoRA management, cinematic resolution controls, sampling, gallery, and local generation.

GPU-accelerated ComfyUI Docker service for the NVIDIA DGX Spark (GB10, aarch64) — two-layer venv, pinned supply chain, self-healing entrypoint

ComfyUI custom nodes for the xAI Grok API: prompt generation, text refinement, vision captioning and batch text loading — with full sampling control (temperature, top_p, penalties, seed-based cache busting) and beginner-friendly tooltips on every option.

SnowLeopard Vision - local AI image and video enhancement tool

Image prompts for Nano Banana (nanobanana), GPT-5 (GPT5), and GPT-4o (GPT4o), including Nanobanana prompts.

an experimental fork of ReActor node with an attempt of making it more self-sufficient aka modern Comfyui environment friendly

infinite-creation 是一个通过提交小说/剧本，本地ComfyUI-MiniMaxH3+Agent+Skill技能全自动无限生产视频，不限时长。

ComfyUI Custom Nodes

ComfyUI node for single-image captioning using frontier multimodal models via OpenRouter and Replicate, with cost estimation and fallback support.

Sweet Tea custom node pack for ComfyUI registry

A ComfyUI node that pauses your workflow and opens a popup window to manually pick which generated images pass through to the rest of the graph.

Custom ComfyUI node that builds MiniMax H3 full-reference prompts.

Wan2GP plugin port of a ComfyUI node for image injection and workflow enhancement. Maintained for Wan2GP integration and usability.

An ultra-fast acceleration node for QwenImage2.1 in ComfyUI.

WindowSeat Reflection Removal with ComfyUI

ComfyUI node for MiniMax H3 that chains multiple video clips with motion context, disk caching, dynamic image references, audio reference support, and seamless final video/audio decoding.

Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4, MiniMax-H3, Gemma 4, FLUX and more.

Visual Version History and Branching for ComfyUI

A ComfyUI custom node for advanced retro pixel-matrix dithering with custom palettes and intelligent color mixing.

Replicates Forge Neo (neo branch) sampling, RNG and prompt encoding inside ComfyUI - same parameters, same results. 在 ComfyUI 中复刻 Forge Neo 的采样流程。

Pre-flight checker for ComfyUI workflows: missing nodes, required models, VRAM feasibility - before installing anything

Qwen-Image 2.1. Checkpoints, quants, prompt engines, LoRAs, and tooling

LoRA Manager for ComfyUI - A powerful extension for organizing, previewing, and integrating LoRA models with metadata and workflow support.

ComfyUI custom nodes for local Ollama LLMs: image prompt generator (spice/fantasy/detail), character prompt builder (AI or template), ACE-Step 1.5 lyrics + tags, and mode-based text rewriter. Edit prompts and dropdowns in JSON—no Python required.

魔芋AI中转API的非官方ComfyUI插件，支持GPT-Image-2、Gemini Banana、Seedream、Seedance、Kling图像与视频生成

🔥 On-Policy Self-Distillation in Diffusion Models

ComfyUI implementation for Normalised Attention Guidance (with Z-Image supported)

ComfyUI node for NovelAI's T5 encoder implementation

Procedural terrain generation with diffusion models (in Minecraft)

Fractal structures grown from a single point, using either cellular automata rules or convolutional kernels.

12 电商出图 Agent Skills for Claude Code / Codex / Cursor · GPT Image 2.5 (Flare/Sunburst) · Amazon 白底主图, A+, 场景图, 模特图, 小红书封面, 淘宝详情页

Local-first AI photo culling desktop app for JPG/PNG + RAW shoots, with OpenAI-compatible model providers and CSV/XMP workflows.

Training-free Spectrum acceleration for ComfyUI’s native MiniMax H3 audio-video model. Uses Chebyshev ridge feature forecasting to skip selected H3 transformer evaluations, with adaptive scheduling, sampler-aware support for Euler, ER-SDE, RES, SEEDS and SA-Solver, CPU/VRAM history storage, and fail

Run Stable Diffusion on Android Devices with Snapdragon NPU acceleration. Also supports CPU/GPU inference.

Official ComfyUI nodes and workflows for ControlFoley: https://github.com/xiaomi-research/controlfoley