ComfyUI is a free, open-source node-based interface for running AI image generation locally on your own hardware. This page brings together every ComfyUI guide on Serverman — whether you are just getting started, trying to install a specific model, or comparing your options. Use the sections below to find what you need.
Getting Started with ComfyUI
If you are new to ComfyUI, start here. These guides cover what it is, whether it is right for you, and how to get it running on your machine.
- What is ComfyUI? Plain-English Guide to Local AI Image Generation — what ComfyUI does, how it compares to other tools, and who it is for
- ComfyUI Beginner’s Guide: Your First Image in 10 Minutes — the node interface explained, loading a model, running the default workflow
- ComfyUI vs Automatic1111: Which Should You Use? — honest comparison of the two main local image generation tools
Installation Guides
Step-by-step installation instructions for every platform. Each guide covers prerequisites, common errors, and first-run setup.
- How to Install ComfyUI on Windows 11 (Step-by-Step Guide) — Python, Git, virtual environment, ComfyUI Manager, and your first image
- How to Install ComfyUI on Mac (Apple Silicon and Intel) — MPS acceleration on Apple Silicon, Intel CPU fallback, performance expectations
Models and Hardware
Choosing the right model and knowing your hardware requirements will save you a lot of trial and error. These guides cover both.
- Best Models for ComfyUI in 2026 — Flux.1, SDXL, SD 1.5 and Pony Diffusion compared by quality, VRAM and use case
- How to Install Flux in ComfyUI (Step-by-Step Guide) — Flux.1 Dev vs Schnell, required files, folder structure, basic workflow
- Best GPU for ComfyUI in 2026: VRAM Requirements and Recommendations — VRAM requirements per model tier, specific card recommendations for UK buyers
Advanced Techniques
Once you have ComfyUI working and have generated your first images, these guides cover the most powerful techniques for improving your results and extending what ComfyUI can do.
- How to Use LoRA in ComfyUI: Step-by-Step Guide — where to put files, the LoRALoader node, strength settings, and stacking multiple LoRAs
- How to Use ControlNet in ComfyUI: Step-by-Step Guide — pose, depth and edge control, preprocessors, and building a controlled workflow
- How to Upscale Images in ComfyUI — RealESRGAN, ESRGAN models, tile upscaling, and combining with img2img for detail enhancement
Custom Nodes and Extensions
Custom nodes are how ComfyUI grows beyond its defaults. This guide covers everything from installing your first node to managing a full library.
- How to Install Custom Nodes in ComfyUI (With and Without Manager) — ComfyUI Manager, manual git clone installs, and the most essential nodes to add first
Linux Installation
- How to Install ComfyUI on Linux (Ubuntu and Debian Guide) — CUDA setup, virtual environments, launch flags, and running headless on a server
Troubleshooting
- ComfyUI Not Working? Common Errors and How to Fix Them — out-of-memory errors, black images, missing models, stuck queues, and red error nodes
About ComfyUI
ComfyUI was created by comfyanonymous and first released in 2023. Unlike Automatic1111, which uses a traditional form-based interface, ComfyUI works as a node graph — each step in the image generation pipeline is a separate node, and you connect them together visually. This makes it more complex to learn but significantly more powerful and flexible once you understand it.
ComfyUI supports Stable Diffusion 1.5, SDXL, and Flux.1 out of the box, with community-built custom nodes adding support for ControlNet, LoRA stacking, upscaling, video generation, and more. The ComfyUI Manager extension makes installing and updating custom nodes straightforward without needing to touch the command line.
All guides on this page cover both Windows and Mac, with Linux steps included where they differ. GPU recommendations focus on NVIDIA CUDA (the most compatible option) with AMD and Apple Silicon coverage included throughout.
What to Read First
If you are completely new, the best order is: read the What is ComfyUI overview first, then follow the installation guide for your OS, then work through the Beginner’s Guide to understand the interface before you start experimenting with different models.
If you already have ComfyUI installed and working, the most useful guides are likely the Best Models for 2026 and the Flux installation guide — Flux.1 is a significant step up in output quality and worth setting up if your hardware supports it.
ComfyUI System Requirements
ComfyUI runs on Windows, Mac and Linux. The minimum hardware depends on which models you want to use:
- SD 1.5 models: 4 GB VRAM minimum, 6 GB recommended. Runs on most modern GPUs.
- SDXL models: 6 GB VRAM minimum, 8 GB recommended for comfortable generation at 1024×1024.
- Flux.1 Schnell (fp8): 8 GB VRAM minimum, 12 GB for smooth operation without quantisation.
- Flux.1 Dev (full precision): 16 GB VRAM recommended. 24 GB for no compromises.
Apple Silicon Macs use unified memory, so an M2 Pro with 16 GB has 16 GB available for ComfyUI. This makes Apple Silicon a genuinely competitive option for Flux.1 Schnell workflows. For a full breakdown of GPU options and UK pricing, see the Best GPU for ComfyUI guide.
ComfyUI vs Other Tools
ComfyUI is the most flexible local image generation tool available, but it is not the easiest to learn. Automatic1111 (A1111) and its fork Forge are better starting points if you just want to generate images quickly without learning a node-based interface. ComfyUI becomes the better choice once you want to build custom pipelines, use ControlNet, chain multiple models together, or work with Flux. The full ComfyUI vs A1111 comparison covers the differences in detail.
