Stock photos look the same everywhere, and hiring a designer for every social post is expensive. ComfyUI lets a small business generate product shots, lifestyle images, and ad creatives on its own computer, with full control over every setting. The trade-off is a steeper learning curve than a one-click app. This tutorial shows you how to install ComfyUI and build your first image workflow.
What It Is
ComfyUI is a free, open-source, node-based interface for running AI image, video, audio, and 3D models locally. Instead of typing a prompt into a single box, you build a workflow out of connected blocks called nodes: one node loads the model, another turns your text into a prompt, another generates the image, and another saves it. The project is published under the GNU General Public License v3.0 and, as checked on October 17, 2026, has 138,022 stars on GitHub. Note that the repository has moved: the old comfyanonymous/ComfyUI address now redirects to github.com/Comfy-Org/ComfyUI. Its official description is: “The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.”
Who It’s For / SMB Use Case
ComfyUI fits businesses that create visual content regularly and want repeatable, brand-consistent results. Practical uses include:
- Product and lifestyle photos: Generate clean product shots, backgrounds, and seasonal variations without booking a photo shoot for every idea.
- Ads and social content: Produce many ad variations from one approved workflow, then pick the winners.
- Repeatable workflows: Save a workflow as a JSON file and reuse it, so every image follows the same steps, sizes, and style.
- Privacy and cost control: Run models locally with no per-image fee, and keep unreleased product images on your own machine.
ComfyUI is not the right pick if you only need one quick image a month. A simple hosted generator will be faster. It earns its place when your team produces visuals weekly and wants control, consistency, and no per-image bill.
Installation
ComfyUI runs on Windows, Linux, and macOS. The desktop app is the easiest path for beginners. All steps below come from the official ComfyUI README.
Option 1: Desktop app (recommended for beginners)
Download the desktop application from comfy.org/download. It is available on Windows and macOS, and the README calls it “the easiest and best way to use ComfyUI for new users.”
Option 2: Portable Windows build
There is a portable standalone build for Windows that works on NVIDIA GPUs or on CPU only. Download the 7z file for your GPU, extract it with 7-Zip or Windows Explorer, and run it. If you have trouble extracting it, right-click the file, open Properties, and click Unblock. The README notes the portable build currently comes with Python 3.13 and PyTorch CUDA 13.0, and advises updating your NVIDIA drivers if it does not start.
Option 3: Install with comfy-cli
You can install and start ComfyUI using comfy-cli:
pip install comfy-cli comfy install
Option 4: Manual install (Windows, Linux)
For a manual install, clone the repository, then place your model files in the right folders: checkpoints go in models/checkpoints and VAEs in models/vae. Python 3.13 is very well supported, according to the README. For NVIDIA GPUs, install PyTorch with:
pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu130
Then install the dependencies from inside the ComfyUI folder:
pip install -r requirements.txt
For AMD, Intel, Apple Silicon, Ascend, Cambricon, and Iluvatar hardware, the README gives separate PyTorch commands. Check the official README for the command that matches your hardware before installing.
How to Use ComfyUI
Step 1: Launch ComfyUI and load a starting workflow
For a manual install, launch ComfyUI by running:
python main.py
Open the web interface in your browser. Start from the default graph, or load a ready-made template from the workflow library. Templates are the fastest way to get a working result before you build your own graph.
Step 2: Understand the default text-to-image graph
A basic workflow has a chain of nodes: a checkpoint loader, a text encoder for the positive prompt, a text encoder for the negative prompt, a sampler that generates the image, a VAE decoder that turns the result into pixels, and a save node. Click a node to change its settings. The prompt nodes are where your business description goes, for example: “minimal product photo of a ceramic coffee mug on a wooden cafe table, soft morning light.”
Step 3: Queue your first image
Press Ctrl + Enter to queue the current graph for generation. The image appears when the run finishes. If you change only the prompt, ComfyUI re-runs just the parts of the graph that changed, which saves time when you are testing variations. Press Ctrl + S to save your workflow.
Step 4: Reuse a workflow from any generated image
ComfyUI saves the full workflow inside supported generated images. Drag a generated PNG back onto the ComfyUI web page to restore the complete workflow, including seeds. This is how a team shares an approved look: send the image file, and anyone can reproduce or tweak it.
Step 5: Add custom nodes with ComfyUI-Manager
ComfyUI-Manager is an extension for installing, updating, and managing custom nodes. Install its dependencies:
pip install -r manager_requirements.txt
Then enable it when launching ComfyUI:
python main.py --enable-manager
Only install custom nodes from sources you trust, and test them on a copy of your workflow first.
Pricing / Limitations
ComfyUI itself is free and open source under the GPL-3.0 license, and it can run fully offline. Your costs are hardware, electricity, and disk space for models, which can be several gigabytes each. The README states ComfyUI can run even the biggest open-source models on as low as 4GB VRAM plus 8GB RAM, but bigger models and video generation will be slow without a strong GPU. Comfy Cloud is the official paid cloud version for people who cannot afford local hardware; check the official site for current cloud pricing. Some partner nodes access closed-source models through paid APIs. The main limitations are the learning curve, the time needed to find good models and settings, and the need to respect each model’s own license before using generated images commercially. If the README does not state something clearly, check the official docs at docs.comfy.org.
Bottom Line for SMB
ComfyUI is the best free option for a small business that wants repeatable, brand-consistent AI visuals without paying per image. Start with the desktop app and a template workflow, get one approved look working, then save it and reuse it. Skip it if you only need an occasional image. For prompt-writing skills that transfer directly to ComfyUI, see our AI prompt templates for small business and our roundup of the best AI writing tools for small business.
Frequently Asked Questions
Do I need a powerful GPU to use ComfyUI?
No, but it helps. The README says ComfyUI can run on as low as 4GB VRAM plus 8GB RAM. Smaller models and images will run on modest hardware, while large models and video need a stronger GPU or the paid cloud version.
Is ComfyUI free for commercial use?
ComfyUI itself is free and open source under the GPL-3.0 license. However, each AI model you download may carry its own license, so check the model’s license before using images in paid advertising.
How is ComfyUI different from a one-click image generator?
One-click tools hide the settings. ComfyUI exposes them as a node graph you can save, share, and reuse, which gives you consistent results and full control at the cost of a longer setup.
