Stable Diffusion on Windows: Your First Image in 20 Minutes

admin · Apr 7, 2026 · 1 views · 5 min read

What You Need

Windows 10 or 11
An NVIDIA GPU with at least 6GB of VRAM. 8GB+ is better. If you have an RTX 3060, 3070, 4060, or anything above — you're good. AMD GPUs work but the setup is harder and slower. If you don't have a GPU, check out the cloud setup guide instead.
At least 20GB of free disk space. Checkpoints are 2-7GB each, and you'll want a few.

That's it. Let's go.

Step 1: Install Python

Download Python 3.10 from python.org. Scroll down to the bottom and grab the Windows installer (64-bit).

When you run the installer, check the box that says "Add Python to PATH" at the bottom of the first screen. This is important — if you miss this, nothing else will work. Then click Install Now.

To verify it worked, open Command Prompt (type cmd in the Start menu) and type:

python --version

You should see Python 3.10.x. If you see an error or a different version, uninstall and try again with that PATH checkbox checked.

Step 2: Install Git

Download Git from git-scm.com. Run the installer and just click Next through everything — the defaults are fine.

Step 3: Download Automatic1111

Open Command Prompt and navigate to where you want to install it. Your desktop or a dedicated folder works fine:

cd C:\Users\YourName\Desktop
git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git

This creates a stable-diffusion-webui folder. It'll take a minute to download.

Step 4: Download a Checkpoint

A checkpoint is the AI model that actually generates images. Without one, the software has nothing to work with.

Here are some good first checkpoints depending on what kind of images you want to make:

For anime/illustration:

Anything V5 — classic anime style, very forgiving with prompts
MeinaMix — clean anime with good anatomy

For realistic/photographic:

Realistic Vision — the go-to for realistic portraits
epiCRealism — natural-looking with good skin

For stylized/3D:

DreamShaper — versatile, does everything from 3D to painterly
RevAnimated — great for fantasy and 3D-style art

Pick one that matches what you want to make. You can always download more later. Click the download button on Civitai to get the .safetensors file.

Once downloaded, move the file to:

stable-diffusion-webui\models\Stable-diffusion\

Step 5: Launch It

Double-click webui-user.bat inside the stable-diffusion-webui folder.

The first launch takes a while — it's downloading dependencies and setting things up. You'll see a lot of text scrolling in the terminal. Don't close it. Wait until you see something like:

Running on local URL: http://127.0.0.1:7860

Open that URL in your browser. You should see the Automatic1111 interface.

Step 6: Your First Image

In the top-left dropdown, select the checkpoint you downloaded. If it doesn't show up, click the refresh button next to the dropdown.

Now type a simple prompt. Don't overthink this — just describe what you want to see:

a girl standing in a field of flowers, sunset, beautiful lighting

Leave negative prompt empty for now.

The only two settings you need to care about right now:

Steps: 20-30. This is how many passes the AI takes to build the image. 20 is fast and decent. 30 is better quality but slower. Don't go above 40 — you won't see much improvement.
Width and Height. Start with 512x768 for portrait or 768x512 for landscape. These are the standard sizes for SD 1.5 checkpoints. If you're using an SDXL checkpoint (the file will usually say SDXL in the name), use 832x1216 for portrait or 1216x832 for landscape.

Leave everything else at defaults. Sampler, CFG, seed — don't touch them yet. They matter, but not for your first image.

Click Generate.

Your first image should appear in 10-30 seconds depending on your GPU. It probably won't be amazing. That's normal. Click Generate again with the same prompt — you'll get a completely different image because the seed changes each time.

Step 7: Play

Now just experiment. Try different prompts. Try different sizes. Try more steps and fewer steps so you can see the difference.

A few things you'll discover quickly:

Size changes the composition, not just the resolution. A tall image (512x768) gives you portraits. A wide image (768x512) gives you landscapes. A square image gives you centered compositions.
More steps = more detail, but with diminishing returns. The jump from 10 to 20 steps is huge. The jump from 30 to 50 is barely noticeable.
The same prompt gives different results every time. That's by design. If you find one you love, write down the seed number (shown below the image) so you can recreate it.

Don't worry about negative prompts, samplers, CFG scale, or any of the advanced settings yet. Just generate images. Get a feel for what your checkpoint can do. You'll learn the rest naturally as you go.

What's Next?

Once you're comfortable generating and you want to start getting better results, check out the rest of the tutorials:

Why I don't start with a prompt — the mindset shift that changes everything
The Three Starting Points — how to go from reference images to better prompts
How I Remix Any Prompt — how to take someone else's prompt and make it your own

Welcome to the rabbit hole.

← Back to Tutorials