Over the past weeks, we’ve seen what Stable Diffusion can do, the new Stability AI artificial intelligence that allows images to be created through a text description. DALL-E 2 was the tool that set a precedent in this regard, and after that, many appeared Being able to try what they are capable of for free. Stable Diffusion is one of them, and while it has become popular in a “weird” way, it is one of the best tools to try this image creation method for free.
Through the DreamStudio or Hugging Face sites, it is possible to test the capabilities of Stable Diffusion for free. However, for those who want to install this language model locally, now You can do it very easily by just downloading a small installer.
Requirements to install Stable Diffusion locally
Running Stable Diffusion locally has many benefits, although most of them are geared towards developers and artists. Install this AI on a file the computer It allows you to explore the possibilities and applications of this tool in all kinds of environments. Many independent developers are adapting this language model in all kinds of applications, and in the future, artificial intelligence such as DALL-E 2 or Stable Diffusion is expected to have a greater role in drawing or 3D modeling tools.
To avoid problems when running Stable Diffusion locally, You’ll need a decent dedicated graphics card And a good amount of VRAM. This will allow the results to be shown in a more flexible and accurate way while there is enough capacity to be able to use the computer in a normal way.
How to install Stable Diffusion on your PC
Once we consider the requirements, we will show you how to install Stable Diffusion locally in a simple way and without having to rely on installing independent environments or libraries, since The download file will do everything for us.
All we have to do is go to the GitHub page for the stable deployment UI. This project is dedicated to focusing everything that is necessary Install Stable Diffusion in a single BAT file Which will download all the necessary components. Installation is available for windows and Linux computers. We just have to click on the link we want.
Once we download the ZIP file, we extract it and We move the folder to the root of our volume. Example: C:Fixed-Propagation-UI. This is recommended so that the installation does not cause problems.
When we move the folder, We just have to open the file “Start Stable Diffusion UI.cmd” And in the command window, we will see the process of downloading all components. Do not give up, because the process can take more than 10 or 15 minutes, even with a computer with good hardware and Internet Connection.
When the process is about to finish, a page will automatically open in our browser with the URL of “https://localhost:9000“. this will be The address we must include in our browser if we want to access the tool. It is an interface where we can insert the description of the image we want to create with some advanced settings. If the “Stable Diffusion is started” message remains for a long time, we can close and reopen the CMD window from the “Start Stable Diffusion UI.cmd” file.
Every time we want to open Stable Diffusion We will have to do this from the same BAT file that we used for the installation, although after downloading and installing everything, we won’t have to wait long for the server to be configured. In the CMD window we can see the process of creating images, while in the web interface we will see the result of our description.
One of the benefits of using Stable Diffusion locally is Be able to save images in high resolution, although viewing images in 4K resolution and it will take a little longer. In addition, from the advanced settings of the web interface, we can also check a box to correct the distorted faces of the images. This works fine, although the refinement is noticeable.
In another tab of the web interface we also find modifiers that we can apply to the image, such as visual style, custom tags, etc.
The tool will automatically update the packages when the BAT file is executed, so you don’t have to worry about anything, just think about what we want to create with the image.