How to install comfyui?
The easiest way to install ComfyUI is the official desktop installer or the portable package from github.com/comfyanonymous/ComfyUI: download, unzip, place a model checkpoint in models/checkpoints, and run the launcher. ComfyUI is a free, node-based Stable Diffusion interface where you connect boxes to build image workflows. It runs locally at 127.0.0.1:8188.
Why — the first-principles explanation
ComfyUI is a node-based interface for Stable Diffusion. Instead of a fixed form with a prompt box, you see a canvas of connected boxes (nodes), each doing one job: load model, encode prompt, sample, decode, save. This graph mirrors exactly how the model actually works, which is why advanced users prefer it: total control and easy reuse of complex pipelines.
Installation is simpler than AUTOMATIC1111 because ComfyUI ships portable builds with Python and dependencies bundled. On Windows you download a zip, unzip it, and run a batch file, no separate Python 3.10.6 install needed. There's also an official desktop app installer for one-click setup, and the manual Git route for those who want it.
Like every Stable Diffusion tool, ComfyUI is just the interface; it needs a model checkpoint to generate anything. You place .safetensors files in the models/checkpoints folder. The same modular folder system holds VAEs, LoRAs, and ControlNet models, each in its own subfolder.
A huge part of ComfyUI's power is the ComfyUI Manager, a community add-on that installs custom nodes and missing models with a few clicks. Because workflows are shareable as JSON or embedded in a PNG, you can drag someone else's example image onto the canvas and instantly load their entire node setup. This 'download a workflow and run it' culture is why ComfyUI spread so fast.
An example that makes it click
AUTOMATIC1111 is like a microwave with preset buttons: fast and simple. ComfyUI is like a set of building blocks where you snap together each part of the machine yourself, load model, add prompt, cook, plate. It takes a minute to learn, but then you can build any contraption you imagine.
And here's the cool part: if a friend builds a great contraption, they hand you a single photo, you drag it onto your screen, and all their building blocks appear already connected. You press go and get the same result.
How to do it
- Go to the official repo at github.com/comfyanonymous/ComfyUI.
- Choose the desktop installer (easiest) or download the portable ZIP for Windows.
- Unzip the portable package to a folder, or run the desktop installer.
- Place a model checkpoint (.safetensors) in the ComfyUI/models/checkpoints folder.
- Run run_nvidia_gpu.bat (NVIDIA) or run_cpu.bat, or launch the desktop app.
- Open the local address (default http://127.0.0.1:8188) and load the default workflow.
- Install ComfyUI Manager to easily add custom nodes and missing models.
Key facts
- The official ComfyUI repository is github.com/comfyanonymous/ComfyUI.
- ComfyUI offers a portable ZIP (Python bundled) and an official desktop installer for one-click setup.
- Model checkpoints go in ComfyUI/models/checkpoints; it runs locally at http://127.0.0.1:8188 by default.
- ComfyUI is node-based, letting users build and reuse custom image-generation workflows.
- Workflows can be shared as JSON or embedded in PNG files and loaded by dragging the image onto the canvas.
The open-source image model you can run on your own hardware.
Affiliate link — we may earn a commission at no cost to you.▶ The 60-second explainer (script)
Let's install ComfyUI, the powerful node-based way to run Stable Diffusion. Instead of a single prompt box, ComfyUI shows connected boxes you wire together, giving you total control. The good news: installing is easy. Go to the official ComfyUI GitHub page. You have two simple options. One, download the desktop installer for a one-click setup. Two, grab the portable ZIP for Windows, which already includes Python, so just unzip it. Either way, you need a model. Drop a .safetensors checkpoint into the folder called models, then checkpoints. Now run the launcher, run_nvidia_gpu.bat if you have an NVIDIA card, and open the local address it shows, usually 127.0.0.1 colon 8188. You'll see a default workflow ready to generate. One last tip: install ComfyUI Manager, a free add-on that lets you install custom nodes and missing models with a click. And because workflows can be saved inside a PNG, you can drag someone's example image onto the canvas to instantly load their whole setup.
What authoritative sources say
People also ask
Is ComfyUI harder than AUTOMATIC1111?
It has a steeper learning curve because of its node graph, but installation is often easier thanks to portable and desktop builds, and it offers more control.
Do I need to install Python separately?
Not for the portable Windows package or desktop app, which bundle Python. Only the manual Git install requires your own Python setup.
Where do models go in ComfyUI?
Checkpoints go in models/checkpoints, LoRAs in models/loras, and VAEs in models/vae, each in its own subfolder.
How do I load someone else's workflow?
Drag their workflow JSON or a PNG image that has the workflow embedded onto the ComfyUI canvas, and the full node setup appears.