Flux 2 KLEIN + ComfyUI Definitive Setup Guide (2026)

Introduction

Flux 2 KLEIN is the definitive model for local image generation in 2026. This guide explains the construction steps in a "portable configuration" by grouping all files into one folder (e.g., C:\flux-klein) to keep your main system environment clean.

Step 0

Prerequisites

Ensure you have the following environment ready:

✅ Python 3.12+ (winget install Python.Python.3.12)
✅ Git (winget install Git.Git)
✅ Storage: 30GB+ free space (SSD recommended)
✅ VRAM: 8GB+ (NVIDIA RTX 3060 or higher)

⚠️ Note: Installation will consume significant C: drive space with temporary files. Beware of Temp capacity shortages.

Step 1

Environment Setup: Building the Foundation

Create a dedicated folder and virtual environment.

                    # Create a folder on any drive
                    mkdir G:\flux-klein
                    cd G:\flux-klein

                    # Create virtual environment (venv)
                    python -m venv venv

                    # Activate virtual environment
                    .\venv\Scripts\Activate.ps1
                

Install PyTorch 2.6

Install PyTorch with CUDA support. If your C: drive space is low, change the temporary folder using the following steps:

                    # (Space shortage fix) Move Temp folder to G: drive
                    $env:TMP = "G:\flux-klein\tmp"; mkdir $env:TMP
                    $env:TEMP = "G:\flux-klein\tmp"

                    # Execute installation
                    pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
                    --no-cache-dir
                

Step 2

ComfyUI & Custom Nodes

Install the UI base and Flux 2 extensions.

                    # Clone ComfyUI
                    git clone https://github.com/comfyanonymous/ComfyUI.git
                    cd ComfyUI

                    # Install dependencies
                    ..\venv\Scripts\pip.exe install -r requirements.txt
                    ..\venv\Scripts\pip.exe install xformers --index-url https://download.pytorch.org/whl/cu121

                    # Flux custom nodes
                    cd custom_nodes
                    git clone https://github.com/kijai/ComfyUI-FluxKleinNodes.git
                    cd ComfyUI-FluxKleinNodes
                    ..\..\..\venv\Scripts\pip.exe install -r requirements.txt
                

Step 3

Model File Placement

Move the four downloaded files to their designated locations.

File Type	Filename	Path (inside ComfyUI)
Main Model	flux-2-klein-4b.safetensors	`models/unet/`
VAE	ae.safetensors	`models/vae/`
CLIP	clip_l.safetensors	`models/clip/`
T5 Checkpoint	t5xxl_fp16.safetensors	`models/clip/`

Step 4

Launch and Run

Create a batch file for easy launching.

                    @echo off
                    call venv\Scripts\Activate.ps1
                    python ComfyUI\main.py --listen 127.0.0.1 --port 8188 --lowvram
                    pause
                

Save the above as launch.bat and run it. Access http://127.0.0.1:8188 in your browser, and you're set!

🎉 Verification: If "FLUX Klein" appears in the node list on the right side of the screen, you've succeeded.

Troubleshooting

Q. "Failed to import" error

Dependencies for custom nodes were not correctly installed. Re-run pip install -r requirements.txt.

Q. Crash due to lack of VRAM (OOM)

You have the --lowvram option enabled, but if it still crashes, close other browser tabs like Chrome to free up VRAM. On 8GB systems, it's a tight fit, so avoid using other apps simultaneously.

🚑 If you still can't get it to work manually?

Ask Antigravity to "build this environment," and it can set everything up in minutes. However, it's best to handle copying or downloading large model files manually.

Read Antigravity Review

Flux 2 KLEIN + ComfyUI Setup

Introduction

Prerequisites

Environment Setup: Building the Foundation

Install PyTorch 2.6

ComfyUI & Custom Nodes

Model File Placement

Launch and Run

Troubleshooting

Q. "Failed to import" error

Q. Crash due to lack of VRAM (OOM)

🚑 If you still can't get it to work manually?

🎨 Check Other Image Models

Introduction

Prerequisites

Environment Setup: Building the Foundation

Install PyTorch 2.6

ComfyUI & Custom Nodes

Model File Placement

Launch and Run

Troubleshooting

Q. "Failed to import" error

Q. Crash due to lack of VRAM (OOM)

🚑 If you still can't get it to work manually?

🎨 Check Other Image Models

Midjourney v7

Grok Image

Adobe Firefly