1-1 Let's create an image with AI(Stable Diffusion Practical Guide Table of Contents)

>>> Let's enjoy AI image creation

Before learning AI image creation, let's first experience AI image creation. If you have never experienced text image creation, the first thing I recommend is the Niji ㆍ Journey website.

https://nijiourney.com/

Niji ㆍ Journey is an AI image creation service released in July 2022. It is a service developed by Spellbruch specializing in creating animation and illustration-style images based on Midjourney, which was popular for a time. The paid subscription type is compatible with this. These two services can be used with the chat service Discord application. In October 2023, a smartphone application version of Niji ㆍ Journey was released, and from March 2024, you can currently create up to 20 images for free. Here, let's create an image using the smartphone application version.

https://play.google.com/store/apps/details?id=com.spellbrush.nijijourney&pli=1

>>>Niji ㆍ Journey: Let's create images with Android and iPhone apps

First, download the application from the App Store, set the language, and the service agreement agreement screen will appear. If you are using it for the first time, it is a good idea to check the usage method by following the application's tutorial.

>>> Let's enjoy AI image generation

-Preparing a Japanese animation-specific version of Focus

You can create images without specialized knowledge using AI image generation services such as Niji and Journey, but you who are reading this article are probably not satisfied with that level. Fooocus, which I'm introducing now, is easy to operate like Midjourney or Niji and Journey, and you can create high-quality text images for free.

Foocus was developed mainly by Lvmin (Lvmin zhang) of Stanford University and is a UI (webUI) that can be used in web browsers. Since it is open source, you can download the program from GitHub. And if you prepare the computing environment yourself, you can use it for free.

GitHub - IIIyasviel/Fooocus

https://github.com/lllyasviel/fooocus

The internals of Focus have Stable Diffusion XL (SDXL), which is covered in this article, and you can obtain very high-quality results even with simple text input. The environment is similar to when using Midjouney and Niji/Journey.

First, log in to your Google Chrome account.

Then, access the URL below to display AICU's GitHub, and click the 'Open in Colab' button on the screen.

Fooocus Japanese Animation Special Edition

https://j.aicu.ai/FoooC

This will cause Colab to appear in the Google Drive you're logged in to, so click 'Copy to Drive' to save the notebook.

Once saved, a new window will automatically open, and you can work on a copy of the notebook in your Google Workspace. Scroll down the notebook and click 'Run cell' under Fooocus JP + Google Drive

output.You will be asked to grant permission to access files on your Google Drive. Click ‘Access Google Drive’ and wait 100for the program to launch.

The program will run and display the code, but wait a little longer until you see the URL https://...gradio.live . Once the URL is displayed, you can click on it to access Fooocus's WebUI in another tab. This URL is valid for up to 72 hours and can be used on smartphones and other PCs.

>>> Let's enjoy AI image generation

--Generate in Focus Japanese animation specialized version

Fooocus is very simple to use. When you open the WebUI, first enter 1girl to the left of the 'Generate!' button, click 'Generate!', and wait for a few seconds. Then, 2 female images will be generated.

Next, enter another prompt and click 'Create!' again. For example, enter ‘prompt: Smiling, 1girl, bluesky, looking at viewer’ and click Create. This will create a new image reflecting what you entered in the prompt.

From now on, I encourage you to explore and investigate the prompt and the WebUI and try it out for yourself. If you check 'Advanced Settings', you will see the extended settings on the right side of the screen, where you can change various settings.

You can choose between 'Speed ​​Priority', 'Image Priority', and 'Ultra-fast' for the performance, and you can change the image quality and generation speed. For example, if you want to make the width longer, you can choose '| 1408x704 | 2:1'. The initial setting for the number of images generated per time is 2, and if you change it to '1', it will generate one image at a time, and if you set it to '10', it will play 10 images continuously, although it will take some time. The negative prompt is to express 'what you do not want to see' in the command window. If you uncheck the random checkbox, similar types of images will come out well.

Next, select the style in the 'Style' tab of the extended settings. One of the features of Fooocus is that you can mix multiple distinctive styles. If you place the mouse pointer over each style, a preview represented by a cat will appear. Using this as a hint, you can check the On/Off checkboxes to find the style you want.

I created an image by selecting SAI Origami, SAI Pixel Art, SAI Line Art, and Artstyle Abstract Expressionism one by one from the top. You can mix styles by checking multiple checkboxes, but I recommend trying one style at a time for now.

In this way, even if you don't know the technology or parameter tuning applied inside, you can still create high-quality text images. However, there may be some people who have difficulty creating the images they want to create. So from now on, we will look into the establishment process of AI technology and the image creation structure in detail so that you can skillfully utilize the image creation AI as you wish.