4 mins read

Who Needs a GPU? Discover the Power of Stable Diffusion using CPU Only with FastSD!

Are you eager to generate stunning images but lacking a powerful GPU? FastSD CPU is the perfect solution for this. This simple tool, based on Latent Consistency Models, allows you to swiftly create high-quality images on your CPU with Stable Diffusion, without the need of a GPU. In this tutorial, we will guide you through the process of setting up and running FastSD CPU on both Windows and Linux systems. Let’s dive in!

Features

Let’s take a quick look at the exciting features FastSD CPU offers:

  • Support for multiple image sizes: 256×256, 512×512, and 768×768
  • Compatibility with both Windows and Linux operating systems
  • Ability to save generated images conveniently
  • Customizable settings for controlling steps, guidance, and seed manipulation

One thing before we get started, you will need git and python installed on your device in order to be able to install and run this tool.

Setting Up FastSD CPU on Windows

For Linux, check the next section.

Step 1: Clone/Download the Repository
First things first, make sure you have the FastSD CPU repository downloaded or cloned onto your local machine.

I used the Terminal for this step; run this command in a folder of your preference:git clone https://github.com/rupeshs/fastsdcpu.git

Step 2: Install FastSD CPU
Go inside the folder where you cloned the repository and double-click on the install.bat file. Sit back and relax while the installation process runs. The duration may vary depending on your internet speed.

Step 3: Start FastSD CPU
Once the installation is complete, simply double-click on the start.bat file, and FastSD CPU will be up and running, ready to create impressive images for you!

During the initial startup, FastSD CPU will download the necessary models, which might take some time depending on your internet speed.

To deepen your understanding and explore more about the used models and the author, check out the following links:

– Project Pages: [https://latent-consistency-models.github.io/](https://latent-consistency-models.github.io/)
– Hugging Face Demos: [https://huggingface.co/spaces/SimianLuo/Latent_Consistency_Model](https://huggingface.co/spaces/SimianLuo/Latent_Consistency_Model)
– Models: [https://huggingface.co/SimianLuo/LCM_Dreamshaper_v7](https://huggingface.co/SimianLuo/LCM_Dreamshaper_v7)

Running FastSD CPU on Linux

In this case, we will use the terminal a bit more.

Step 1: Clone/Download the Repository
Similar to the Windows setup, ensure you have the FastSD CPU repository cloned or downloaded on your Linux system.

Step 2: Navigate to FastSD CPU Directory
Open your terminal and navigate to the directory where you have the FastSD CPU files stored.

Step 3: Set Permissions for Installation Script
Execute the following command in your terminal: chmod +x install.sh

Step 4: Run the Installation Script
Run the installation script by typing: ./install.sh

Step 5: Start FastSD CPU
Set permissions for the start script using: chmod +x start.sh

Then, execute the start script: ./start.sh

Using the tool

This is what you should expect when the application starts: it will open a terminal, where you can keep track of the execution process, and a very simple GUI to add the prompt and set some parameters.

FastSD UI Windows
FastSD first opening

In my case, the interface is rather small and it cannot be increased, so I check directly the output folder to see the images.

Write a prompt and click on generate, and you will see logs in the terminal about the image generation. It took around 50 seconds to generate a 768×768 image, using only my CPU. In particular, I have a 12th Gen Intel(R) Core(TM) i7–12700F 2.10 GHz processor and 16 GB of RAM.


In conclusion, while it’s undeniable that more demanding or creative tasks, like running Stable Diffusion XL, may benefit from the power of a dedicated GPU, FastSD CPU offers an amazing opportunity if you don’t have access to a powerful GPU or any GPU at all. This tool allows even more people to dive into the creativity of Stable Diffusion and create breath taking visuals using the processing might of the CPU alone.

Link to repo here.

One thought on “Who Needs a GPU? Discover the Power of Stable Diffusion using CPU Only with FastSD!

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.