From nothing to your first image · No coding · No Python
This page sets up everything from zero. If your computer has nothing installed yet, start here and follow the steps in order. When you are finished, read the How to Use guide to learn the buttons.
Read this first — it will calm your nerves
You do not need to know how to code.
You do not need to install Python. The program we use already includes it.
You do not need to use the command line or type commands.
This is a one-time setup. You do it once, then you just make pictures.
Do the steps in order, one at a time. You cannot break your computer by following them.
Does my computer qualify?
Before you start, check these. It is fine if you are not sure — you can just try.
A Windows PC. This guide is for Windows.
An NVIDIA graphics card is strongly recommended. The graphics card is what draws the images. If you have one, look for how much VRAM (video memory) it has: 6 GB works, 8 GB or more is comfortable, more is faster and lets you make bigger images.
About 15–30 GB of free space on your drive for the program and the models.
No NVIDIA card? It can still work on some AMD cards or on the processor alone, but it is slower and the setup is different. This guide covers the common NVIDIA path. If you are unsure what you have, that is okay — try it, and if it will not run, that is your answer.
Part 1 — Install ComfyUI (the engine)
ComfyUI is the free program that actually draws the pictures. Comfy Console is the simple control panel for it. You install ComfyUI once.
Get a program that opens zip files. The download comes in a compressed file. Install the free 7-Zip so you can open it. (If you already have 7-Zip or WinRAR, skip this.)
Download ComfyUI Portable. Go to the official ComfyUI releases page. Find the newest release at the top and download the file whose name includes windows_portable (it ends in .7z). It is a large file — let it finish.
Unzip it. Right-click the downloaded file → 7-Zip → Extract Here (or Extract to…). This creates a folder named ComfyUI_windows_portable.
Move the folder somewhere with space. Drag the ComfyUI_windows_portable folder to a drive with room, for example directly onto your C:\ drive. Remember where you put it.
Test that it runs. Open the ComfyUI_windows_portable folder and double-click run_nvidia_gpu.bat.
What you will see A black window opens and fills with white text. This is normal. It is not errors. After a moment it will say it is running and may open a page in your browser. That means ComfyUI works. You can close the black window for now.
Good. The hardest part is done. ComfyUI is installed. Everything after this is smaller.
Part 2 — Add one model (the “brain”)
A model (also called a checkpoint) is the brain that knows how to draw. Without one, no pictures can be made. You only need one to start.
Download one checkpoint. A good all-round choice for anime-style art is a WAI Illustrious checkpoint. You can find checkpoints on Civitai — search for the name, and download the file that ends in .safetensors.
Put the file in the checkpoints folder. Move the .safetensors file into this folder (inside where you unzipped ComfyUI):
What you will see Nothing changes yet — you are just placing a file. The Console will let you pick it later.
Part 3 — Make the launch file
Comfy Console talks to ComfyUI. For that to work, ComfyUI has to be started with one setting turned on (it is called CORS). You make a small launch file once, and use it from then on.
Open Notepad. Press the Windows key, type Notepad, and open it.
Copy this text into Notepad exactly:
.\python_embeded\python.exe -s ComfyUI\main.py --enable-cors-header "*" --preview-method auto
pause
Save it into the ComfyUI folder. In Notepad choose File → Save As. Go into your ComfyUI_windows_portable folder (the same place as run_nvidia_gpu.bat). In the “Save as type” box choose All Files. Name it exactly: run_comfy_console.bat and press Save.
From now on, start ComfyUI with this file. Double-click run_comfy_console.bat whenever you want to use Comfy Console. It starts ComfyUI with the connect setting turned on.
Staying up to date Comfy Console has an ⤓ Update button in the top-right. When a newer version exists, that button lights up — click it and it downloads the new file to your Downloads folder. Then close the app, replace your old comfy-console.html with the new one, and open it again. That is the whole update.
Part 5 — Add the advanced features (optional)
You do not need this to make pictures. Add these only when you want the extra buttons: fixing faces and hands, copying a pose, making pictures bigger with detail, or erasing objects. Comfy Console hides any feature you have not installed, so nothing will look broken if you skip this.
Step A — Get ComfyUI Manager
ComfyUI Manager is a helper that installs add-ons with a click. Many recent ComfyUI versions already include it — look for a Manager button in the ComfyUI window. If you do not see one, follow the simple install steps on the ComfyUI Manager page.
Step B — Install the add-ons (nodes)
In the ComfyUI window, click Manager → Custom Nodes Manager. Search for each of these by name and click Install. Install them one at a time.
The links go to each add-on’s page for reference. The easiest way to install is still through the Manager’s search box — you rarely need the links.
When it is done, fully close ComfyUI and start it again with your run_comfy_console.bat. Add-ons only load when ComfyUI starts.
Step C — Add the extra models
Some features need an extra model file. Download each and place it in the listed folder (all folders are inside ComfyUI_windows_portable\ComfyUI\models). Skip any feature you do not want.
The detector models download by themselves the first time you use the feature (or from Manager’s Model list)
models\ultralytics\bbox
Tip — the easy way ComfyUI Manager also has a Model Manager (a searchable list of models it downloads and files for you). Where you can, use that instead of downloading and moving files by hand.
After adding models, restart ComfyUI and refresh Comfy Console (hold Ctrl and Shift and press R). The new features appear on their own.
Part 6 — Train your own character (optional, advanced)
Totally optional. This adds the 🎓 Train tab, which teaches the tool a specific character from a set of pictures (this is called making a “LoRA”). Skip it unless you want that — making pictures does not need it.
Training is a big, heavy job, so it needs a one-time setup. We made it as simple as we could: one file you download and double-click. It does the rest.
Before you start, you need
An NVIDIA graphics card (the same kind ComfyUI needs).
Your browser may warn about a .bat file — that is normal for an installer. Choose Keep.
Double-click the file you downloaded. A blue window opens and explains what it will do. Read it, then press a key to begin. Nothing happens until you do.
Answer two easy questions when it asks:
Where is your ComfyUI? Paste the folder that has run_nvidia_gpu.bat in it (for example C:\ComfyUI_windows_portable) and press Enter.
Which model? It shows your checkpoints; type the number of the one to learn from (any SDXL / Illustrious one).
Wait. It downloads and builds everything — about 15–25 minutes, mostly waiting. You can leave it alone. When it says ALL DONE, close it.
Restart ComfyUI (fully close it, open it again) and reload Comfy Console. A new 🎓 Train tab appears in the top bar.
How to train, once it is set up Open the 🎓 Train tab. Drop in 15–25 pictures of one character (any size — it crops and describes them for you). Type a one-word name (you will type this word in prompts later to summon the character). Press Train LoRA ▸ and wait ~30–60 minutes. When it finishes, your character appears in the LoRA list — add it, set strength to about 0.8, and put your word in the prompt.
The one thing that matters is the pictures. Use a consistent set — clearly the same character, in varied poses, outfits, and angles, all reasonably clear. Good pictures make a great character; blurry or mixed pictures make a weak one. Everything else is automatic.
Good to know
• You cannot make pictures while training — it uses the whole graphics card. That is normal.
• The Train tab only shows up after the installer runs and ComfyUI is restarted. If you do not see it, restart ComfyUI once more.
• It installs to C:\lora-trainer. Leave it there.
If something does not work
The black window closes instantly / shows a red error Usually your graphics card is not NVIDIA, or a file did not download fully. Try downloading ComfyUI Portable again. If you do not have an NVIDIA card, this path will not run.
The Console dot stays red ComfyUI is not running, or you did not start it with run_comfy_console.bat. Start ComfyUI with that file, wait for it to load, then refresh the Console page.
No models show in the Checkpoint box The .safetensors file is not in models\checkpoints, or ComfyUI was started before you added it. Put the file there, restart ComfyUI, and refresh the Console.
I installed add-ons but the features do not show ComfyUI only loads add-ons when it starts. Fully close it and open it again with your launch file, then refresh the Console.
Remember You cannot break anything by clicking. Restarting ComfyUI and refreshing the Console fixes almost every problem.
