Apps & Guides

Easy Diffusion UI

Easy Diffusion UI is an open source software available for download on GitHub. Here’s how to install it on Ubuntu 22.04 LTS. If you’ve just rented a server, install the GPU drivers and extend your home directory. Then, download the latest release of Easy Diffusion UI:

wget https://github.com/cmdr2/stable-diffusion-ui/releases/latest/download/Easy-Diffusion-Linux.zip

Unpack the downloaded ZIP-archive:

unzip Easy-Diffusion-Linux.zip

Change directory to easy-diffusion:

cd easy-diffusion

Start the installation:

./start.sh

This is a script collection that automatically downloads and installs all necessary components. It also downloads the standard Stable Diffusion model in SafeTensors format. Once all downloads and installations are complete, the Easy Diffusion UI will launch automatically.

Using

The previous article, Stable Diffusion WebUI, outlines a method for accepting connections from the public internet and provides simple login and password authorization. In this case, we aim to demonstrate another universal method for forwarding ports through an SSH connection. We use PuTTY to establish a secure connection to the remote server. You can find more information about it in our guide Connect to a Linux server.

To choose which ports to forward, please open Connection > SSH > Tunnels in the left option tree. Type 9000 in the Source Port field and 127.0.0.1:9000 in the Destination field. Then click the Add button:

After that, you can return to Session and save it for later use. Connect to the remote server as usual. Now, all data that you send or receive at port 9000 on the loopback address 127.0.0.1 will be redirected to the remote server. This method creates a virtual secure tunnel that exists as long as the connection does.

Once Easy Diffusion UI starts and port forwarding is on, you can open a web browser and navigate to the address http://127.0.0.1:9000. We recommend downloading and installing custom models, as described in this article, instead of relying solely on the standard model to generate images. Don’t forget to increase the number of inference steps and adjust the desired image resolution (marked with asterisks).

One of the major benefits of the Easy Diffusion UI is its support for multiple GPUs. When you want to create a batch of images, you can choose how many images will be created in parallel. For example, if you have dual GPU configuration:

You can display the GPU’s load during the image generation process. Establish another SSH connection and execute a single command:

watch -n 1 nvidia-smi

Also, Easy Diffusion UI simplifies your prompts creation as it provides numerous examples of image modifiers. You can mix them to achieve more accurate results:

It’s a good idea to explore PromptBook by OpenArt. This guide can significantly enhance your prompt creation skills. With the Easy Diffusion UI, once the image is generated, you can download it, use as example for generating the next image, or make modifications with just one click:

The most common use of the Upscale button is to increase an image’s resolution. The generative neural network uses the original image as a basis and adds additional pixels, thereby interpolating the source image to the desired size.

When generating faces, issues may arise such as misaligned eyes, disproportionate sizes, or malformed parts. Fortunately, these problems can be solved using the Fix Faces button. Additionally, negative prompts may be utilized to prevent incorrect faces from being generated.

Uninstall

All files, scripts, libraries, and models are stored in a single directory. If you want to remove Easy Diffusion UI from your server, just delete this directory along with all the content:

sudo rm -rf easy-diffusion