In this guide you will learn:

→Install ComfyUI on Windows from scratch
→Configure Python, CUDA, and PyTorch correctly
→Launch with GPU-optimized flags for your VRAM
→Install and manage custom nodes via ComfyUI Manager
→Run your first SDXL workflow in under 30 minutes
→Read benchmarks for RTX 5080, 4090, and 3080

Setup

ComfyUI Complete Setup: RTX 5080 Edition

ComfyUI is the most powerful node-based interface for running local AI image and video generation. This guide gets you from zero to a working workflow on Windows with an NVIDIA GPU — tested live on an RTX 5080 and RTX 3080 16GB.

:::stats :::stat 16GB | VRAM Required :::stat 12 min | Setup Time :::stat 3.2s | SDXL per image :::stat Free | Open Source :::

#Hardware Requirements

GPU	VRAM	Best Use Case
RTX 5080	16GB	Full quality, all models, fast batch
RTX 4090	24GB	Full quality, FLUX FP16, large batch
RTX 3080 16GB	16GB	Full quality, slower on large models
RTX 3080 10GB	10GB	Reduced resolution, GGUF models
RTX 3060	12GB	Standard models, SD1.5 and SDXL
GTX 1660 Ti	6GB	SD1.5 only, very slow

:::note Minimum Requirements 6GB VRAM for SD1.5. FLUX and LTX Video require at least 12GB. For best results with modern models, 16GB is the sweet spot. :::

Hardware Partner

Running these workflows? ComputeAtlas.ai helps you find the right GPU

Optimization is only half the battle. Get precise VRAM benchmarks and hardware recommendations tailored for ComfyUI.

Check GPU Prices →

[AD VISUAL: GPU BENCHMARKS]

#Step 1 — Install Prerequisites

You need three things before cloning ComfyUI: Python 3.10, Git, and the CUDA Toolkit.

Python 3.10.x

Download from python.org — use exactly 3.10.x, not 3.11 or 3.12. Many custom nodes have dependency conflicts with newer versions.

comfyui-workflow.json

# Verify your Python version after install
python --version
# Expected: Python 3.10.11

:::warning Version matters Do NOT install Python 3.11 or 3.12. Custom nodes like AnimateDiff and VideoHelperSuite require 3.10.x. Using a newer version will cause cryptic import errors. :::

Git

comfyui-workflow.json

git --version
# Expected: git version 2.x.x

CUDA Toolkit

Download CUDA 12.1 from nvidia.com/cuda-downloads. Match your driver version.

comfyui-workflow.json

nvcc --version
# Expected: release 12.1

#Step 2 — Clone ComfyUI

comfyui-workflow.json

cd C:\
git clone https://github.com/comfyanonymous/ComfyUI
cd ComfyUI

:::tip Keep it at C:\ComfyUI Installing at the root avoids Windows path length issues that break certain custom node installs. Deeply nested paths cause silent failures. :::

#Step 3 — Install Dependencies

comfyui-workflow.json

# Create virtual environment
python -m venv venv

# Activate it (run this every time you open a new terminal)
.\venv\Scripts\activate

# Install PyTorch for CUDA 12.1
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

# Install ComfyUI requirements
pip install -r requirements.txt

The PyTorch install will download ~2.5GB. This is the step most people mess up — make sure you're using the cu121 index URL, not the default PyPI version.

#Step 4 — Install ComfyUI Manager

ComfyUI Manager adds a one-click node installer directly inside the UI. It's essential.

comfyui-workflow.json

cd custom_nodes
git clone https://github.com/ltdrdata/ComfyUI-Manager
cd ..

After this, every time you find a workflow that needs missing nodes, ComfyUI Manager will detect and install them automatically.

:::pro Install Order Install Manager BEFORE downloading any models. When you load a workflow that requires missing custom nodes, Manager will prompt you to install them all at once. :::

#Step 5 — Download Your First Model

Place models in ComfyUI/models/checkpoints/

Recommended starter:

comfyui-workflow.json

# SDXL Base 1.0 — best quality/speed balance for beginners
# Download from: huggingface.co/stabilityai/stable-diffusion-xl-base-1.0
# File: sd_xl_base_1.0.safetensors (6.5GB)

For FLUX (requires 12GB+ VRAM):

comfyui-workflow.json

# FLUX Dev FP8 — best quality at reduced VRAM
# Download from: huggingface.co/lllyasviel/flux1-dev-fp8
# File: flux1-dev-fp8.safetensors (12.1GB)

#Step 6 — Launch ComfyUI

comfyui-workflow.json

# Standard launch
python main.py

# RTX 5080 / 4090 optimized
python main.py --gpu-only --highvram

# RTX 3080 16GB
python main.py --gpu-only

# Under 8GB VRAM
python main.py --lowvram

Open your browser to: http://127.0.0.1:8188

:::note Bookmark this URL ComfyUI runs as a local server. Bookmark 127.0.0.1:8188 — it's faster than typing it each time. :::

#RTX 5080 Optimal Settings

These settings max out quality on 16GB VRAM with headroom for LoRAs:

comfyui-workflow.json

Model:       FLUX Dev FP8 or SDXL
Batch size:  2–4 images
Resolution:  1024x1024 (SDXL) / 768x512 (video)
Steps:       20–30
CFG Scale:   7.0 (images) / 3.0 (video)
Sampler:     dpmpp_2m
Scheduler:   karras
VAE:         Built-in (baked into checkpoint)
Precision:   fp16

#Installing Essential Custom Nodes

In ComfyUI Manager, search and install these first:

→ComfyUI-Impact-Pack — face detailing, segmentation, upscaling
→ComfyUI_IPAdapter_plus — image prompt control
→ComfyUI-AnimateDiff-Evolved — video animation
→ComfyUI-VideoHelperSuite — video input/output (required for LTX)
→rgthree-comfy — better node organization and groups

#Your First Workflow

→Click Load in the top menu
→Select default_workflow.json
→In the Load Checkpoint node, select your downloaded model
→Click Queue Prompt

You should see your first image in 15–30 seconds on an RTX 5080.

:::tip Speed up iteration While testing prompts, use batch size 1 and 15 steps. Once you find a direction you like, bump to 25 steps and batch 4 for the final run. :::

#Performance Benchmarks

Tested on RTX 5080 16GB, --gpu-only --highvram:

Model	Resolution	Steps	Time
SDXL FP16	1024×1024	20	~3.2s
FLUX Dev FP8	1024×1024	20	~8.1s
SD 1.5	512×512	20	~0.8s
LTX Video 2.3	768×512 97f	25	~5.4s

:::stats :::stat 3.2s | SDXL per image :::stat 8.1s | FLUX Dev FP8 :::stat 5.4s | LTX Video clip :::stat 0.8s | SD 1.5 image :::

#Common Issues

"CUDA out of memory" — Add --lowvram to launch command, reduce batch size to 1, or switch to a GGUF quantized model.

Black images — Your VAE doesn't match the model. Download the correct VAE for your checkpoint and load it manually with a VAELoader node.

Slow generation — Check --gpu-only flag is set and your venv is activated. Without the venv, you'll use the system Python and possibly CPU inference.

Nodes missing — Open ComfyUI Manager → click "Install Missing Custom Nodes". This auto-detects and installs everything a workflow needs.

You're ready to run your first workflow. Next up — training your own LoRA to lock a character or style into any model.

Hardware Partner

Running these workflows? ComputeAtlas.ai helps you find the right GPU

Optimization is only half the battle. Get precise VRAM benchmarks and hardware recommendations tailored for ComfyUI.

Check GPU Prices →

[AD VISUAL: GPU BENCHMARKS]