How to Run Stable Diffusion Locally with Stable Diffusion WebUI Forge

You wanted to make a custom birthday card — something better than the generic ones from the shop. Or you needed a quick illustration for a work presentation, and you didn't want to pay for yet another monthly subscription. The idea of typing a sentence and getting a picture back sounds perfect. Then someone mentioned Stable Diffusion, and you saw "Python", "GitHub" and "CUDA" in the same breath. That's usually where curious people quietly close the tab.

Here's the good news: it doesn't have to be that way anymore. Stable Diffusion WebUI Forge wraps the model into a normal-looking installer. You double-click it, answer a couple of questions, and a friendly window opens.

What you need before you start

Before you touch the installer, gather these:

• A Windows, macOS, or Linux computer made in the last 4-5 years. Older machines will work but run slowly.
• A dedicated graphics card (NVIDIA works best; AMD and Apple Silicon also work, with longer generation times). If you're on a laptop with no separate graphics chip, you can still try it — it just runs much slower.
• At least 10 GB of free disk space for the program plus one starter model.
• A Hugging Face account (free — sign up at huggingface.co). You'll need it to download a model file in Step 3.
• About 30-45 minutes for the first install. After that, it starts in under a minute.

Step 1 — Download the Forge installer

Open your web browser and go to the official Forge GitHub page. Look for the green "Code" button, then click "Download ZIP". The file is roughly 1-2 GB.

You'll see a folder called stable-diffusion-webui-forge appear in your Downloads folder once it's done.

💬 Example path: C:\Users\YourName\Downloads\stable-diffusion-webui-forge

You'll know it worked when you can open the folder and see files like webui.bat (Windows) or webui.sh (macOS/Linux) inside.

Step 2 — Run the one-click launcher

Inside the folder, double-click the launcher file that matches your system: webui.bat on Windows, webui.sh on macOS and Linux. A black terminal window will pop up — don't panic. This is normal. It will download a few extra bits on first run (about 2-3 GB).

Wait until you see a line that says something like Running on local URL: http://127.0.0.1:7860. The first launch takes longer than later ones.

💬 Example line in the terminal: Startup time: 34.2s (includes model loading)

You'll know it worked when your web browser opens automatically to a page titled "Stable Diffusion WebUI Forge" with a text box at the top.

Step 3 — Download a starter model from Hugging Face

The Forge window can generate images, but it needs a model first — the trained "brain" that knows what things look like. Go to Hugging Face, sign in, and search for a beginner-friendly model like "Stable Diffusion 1.5" or "Realistic Vision". On the model's page, click the "Files and versions" tab, then download the largest .safetensors file.

Move that file into the models/Stable-diffusion folder inside your Forge installation.

💬 Example model filename: realisticVisionV51_v51VAE.safetensors

You'll know it worked when the file appears in the folder and you can select its name from the dropdown menu at the top-left of the Forge window (refresh the page if it doesn't show up).

Step 4 — Write your first prompt and generate

In the top-left text box labeled "Prompt", type a description of the image you want. Be specific — mention the subject, the mood, the lighting, and the style. Leave the other settings on their defaults for now. Click the orange "Generate" button on the right.

The first image takes 30-60 seconds (the program is warming up). After that, new images appear in 5-15 seconds.

💬 Example prompt to copy:

"a red apple on a wooden table, soft morning light, photo realistic, shallow depth of field"

You'll know it worked when an image appears in the right-hand panel showing your described scene. If you see an error instead, save the error text — the Common mistakes section below covers the most common ones.

Step 5 — Save and organize your images

Forge places finished images inside an outputs folder inside your installation directory. They're sorted by date and given a timestamped filename. To keep things tidy, create a folder called my-images somewhere easy to find (like your Desktop) and move your favorites there.

The image files are standard .png files — open them in any photo viewer, drop them into a Word document, or share them like any other picture.

💬 Example output path: stable-diffusion-webui-forge\outputs\txt2img-images\2026-08-03\00001-1234567890.png

You'll know it worked when you can open the .png file in your normal photo viewer and see the image at full size.

Common mistakes

• "I get a 'CUDA out of memory' error." Your graphics card doesn't have enough VRAM (the dedicated memory on the GPU). Fix: close Chrome, Discord, and anything else using the GPU, or switch to a smaller model (look for ones labeled "SD 1.5" instead of "SDXL").
• "The Forge window doesn't open my browser." The web interface is still running — manually open your browser and type http://127.0.0.1:7860 into the address bar. The page will appear.
• "My images look weird or distorted." Your prompt is probably too vague or uses conflicting style words. Fix: simplify the prompt, stick to one style at a time, and add quality words like "high detail, sharp focus" at the end.
• "The model dropdown is empty." You put the .safetensors file in the wrong folder. Fix: move it specifically into models/Stable-diffusion, then refresh the Forge page in your browser.

Wrap-up

Running Stable Diffusion locally used to be a weekend project full of error messages. Today, with Stable Diffusion WebUI Forge, it's closer to installing any other desktop program. Your images stay on your computer, there's no monthly fee, and you control every knob from a normal-looking window. Pick a quiet evening this week, follow the five steps above, and try your first prompt. The first image will look rough — the tenth will surprise you.