Super Simple Stable Diffusion Setup

For Windows/Nvidia PCs

Installing Forge

Stable Diffusion is the technology used for AI Image Generation, but there are many different user interfaces available. The easiest one to install is Forge.

  1. Go to https://github.com/lllyasviel/stable-diffusion-webui-forge
  2. Click the >>> Click Here to Download One-Click Package (CUDA 12.1 + Pytorch 2.3.1) <<< link to download the zipped file with everything you need

Screenshot of Forge Download Page

  1. Copy the zipped file to wherever you want and extract it there. You might want to get 7-zip if you don't have it ( https://www.7-zip.org/ )
  2. You should see the files below. .bat files are Windows batch files containing scripts to execute. Double-click update.bat to run it and update everything to the latest version of Forge.

Screenshot of Extracted Files

  1. Double-click run.bat to run Forge. The first time it will automatically install everything you need, which might take a while. Whenever you run it in the future it should start faster. Congratulations, you're done with all the setup! But your AI has no brain.

Downloading a Model

A model is basically the AI's "brain" and contains all the knowledge of styles, characters, concepts, etc. Different models have different knowledge. Some are anime-focused, some realistic, etc. Some places to find models online include:

https://huggingface.co/
https://civitai.com/
https://tensor.art/

For this demo we'll use Nova Furry XL from Civitai: https://civitai.com/models/503815/nova-furry-xl

  1. Click the download link. You may need to may a free CivitAI account if you don't have one.

Screenshot of CivitAI Download Page

  1. It will download a .safetensors file. Put it in the webui/models/stable-diffusion folder. There should be a text file called "Put Stable Diffusion checkpoints here.txt" to indicate you're in the right spot.

Screenshot of Model File Location

  1. Restart Forge by running run.bat again. Now everything should be ready for generating AI images!

Generating AI Images

When you run run.bat it should open a Windows Console window and an Internet Browser window. The Console window is where the program is running and the Internet Browser is how you'll interact with the program. Both need to be open for Forge to work.

Screenshot of Windows
Things might look different depending on the version or the browser so don't worry if you see differences

  1. To generate an image, first select a model from the dropdown in the upper left corner. If you only have one model it should be selected automatically.

Screenshot of Model Selection

  1. Type what you want to see in the Positive Prompt box. A good format is style, character, pose, details, then everything else. Prompts are processed left to right, so put whatever is most important to you first. This example uses: anime screenshot, fox, female, big breasts, sexy, seductive, looking at viewer, bedroom, anthro, furry, solo, masterpiece, best quality, detailed background,

Screenshot of Positive Prompt

  1. Type what you don't want to see in the Negative Prompt box. This box tells the AI what to avoid. This example uses: worst quality, worst aesthetic, lowres, bad anatomy, bad hands, simple background, signature, logo, watermark, Patreon, Twitter, artist name, url, copyright, censored,

Screenshot of Negative Prompt

  1. The other settings should be fine with default values. Here's a brief explanation of the important ones if you want to change them:
  • Sampling and Schedule Type control how the AI handles randomness. You can read more about them at https://stable-diffusion-art.com/samplers/ . Euler a is a popular one because it's fast and makes smoother images, but others can be better for more detail or more realistic images.
  • Sampling Steps is how many iterations the AI uses to make the image. If the image looks noisy you can try increasing the steps. More steps takes longer but can improve quality, but only to a certain point.
  • Width and Height control the size of the image generated. Larger images take more time to generate.
  • CFG Scale stands for Classifier-Free Guidance scale. It controls how closely the AI follows the prompt, with higher values giving more freedom and lower values sticking closer to the prompt. However it mostly seems to affect color saturation, with higher values have more color and lower values more muted.
  • Seed is the value used to create the noise for the generation, with -1 being random. You can click the recycling icon to re-use a seed if you want to re-create an image or tweak the prompt slightly while keeping things mostly the same. You can click the dice icon to reset it to the random -1 value.

Screenshot of Settings

  1. Click the big orange Generate button to generate an image.

Screenshot of Generate Button

  1. Congratulations! You can now generate infinite AI images. They'll automatically be saved to the webui/outputs/txt2img-images folder. Best of luck with all your gens!

Screenshot of Finished Image

Edit

Pub: 29 Mar 2025 21:59 UTC

Edit: 03 May 2025 00:40 UTC

Views: 542