How to Install ComfyUI on PC 2026 – Desktop, Portable & Manual

ComfyUI is the most powerful open-source interface for AI image and video generation. This guide shows three ways to install ComfyUI on your PC in 2026, how to configure it, and how to fix common problems.

Table of Contents

System Requirements

	Minimum	Recommended
GPU	NVIDIA 6–8 GB (or AMD / Apple Silicon)	NVIDIA RTX 16–24 GB
RAM	16 GB	32 GB+
Disk	20 GB SSD	100 GB+ NVMe
Python (manual install)	3.10+	3.12

Method 1: ComfyUI Desktop

Download ComfyUI Desktop from comfy.org.
Run the installer and select your GPU type.
Choose an install folder with plenty of space for models.
Launch it and load a template workflow.

Method 2: Windows Portable Build

Download the portable 7z package for NVIDIA from the ComfyUI GitHub releases page.
Extract it with 7-Zip.
Put model files in ComfyUI\models\checkpoints.
Run run_nvidia_gpu.bat and open http://127.0.0.1:8188.

Method 3: Manual Install (Windows, Linux, macOS)

git clone https://github.com/comfyanonymous/ComfyUI
cd ComfyUI
python -m venv venv
venv\Scripts\activate          # Windows  (source venv/bin/activate on Linux/macOS)
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128
pip install -r requirements.txt
python main.py

Configure ComfyUI

Install ComfyUI-Manager to add custom nodes and missing models from the UI.
Use --lowvram on 6–8 GB cards.
Share model folders with other UIs using extra_model_paths.yaml.

Troubleshooting

“Torch not compiled with CUDA enabled”: reinstall PyTorch with the CUDA index URL above.
Out of memory: use FP8 or GGUF-quantized models, lower the resolution, or add --lowvram.
Missing nodes: open ComfyUI-Manager → Install Missing Custom Nodes.

Run ComfyUI on a Cloud GPU

No powerful GPU at home? You can run ComfyUI on a rented GPU server instead. RunPod’s ComfyUI template starts an RTX 4090 for $0.34/hour. For a fixed monthly price, GPU Mart (RTX A4000 VPS from $119/month, RTX 4090 $409/month) and HOSTKEY (RTX 4090 from €279/month) offer dedicated servers you can keep running 24/7.

Compare providers in our ComfyUI hosting guide.

FAQ

Is ComfyUI free?

Yes, ComfyUI is free and open source.

Does ComfyUI work on AMD or Mac?

Yes: AMD via ROCm (Linux) or DirectML, and Apple Silicon via Metal (MPS).

How much VRAM do I need for FLUX in ComfyUI?

8–13 GB for FLUX.2 [klein]; 24 GB or more for FLUX.2 [dev].

Conclusion

ComfyUI Desktop is the quickest way to start, the portable build suits Windows power users, and the manual install gives full control. If your PC is not powerful enough, host ComfyUI on a cloud GPU.