How do you create 'AI illustrations'? Introducing the differences between various creation methods!

Recently, when looking at social media or the internet, you've probably seen 'AI illustrations' much more often, haven't you?

I think many people already know that 'you can create illustrations with AI'.

But when you actually try to look into how to make them yourself...

'Wait, how do you actually create AI illustrations?'

There are methods like asking an AI such as Google Gemini or ChatGPT to create them for you, using image generation services, or setting up an environment on your own computer to generate them; the more you look into it, the more options you find.

Moreover, the way they are used, their difficulty, and what they can do are quite different.

I myself have tried various methods since I became interested in AI illustrations, but at first, I didn't really understand how they differed.

So, this time,

'I want to try making AI illustrations, but what methods are there?'

For those who are wondering this, I would like to introduce the main ways to create AI illustrations and their differences from a beginner's perspective!

1. What exactly is an AI illustration?

First, let's briefly look at what 'AI illustration' actually is.

As the name suggests, an AI illustration isan illustration or image created using AI (artificial intelligence).

For example,

'An illustration of a cat taking a nap under a blue sky'

If you describe the content of the image you want to create in text like this, the AI will generate an image based on that content.

Recently, you can do much more than just use text; you can load a source image to transform it into a different illustration, or modify only a part of a generated image.

Hearing this, you might think,

'So, if I just tell the AI what I want to create, I can make it easily!'

you might think that.

Of course, there are ways to create them just like that.

However, when you actually look into AI illustrations, you will find that depending on the service used and the generation method,there are significant differences in ease of use, flexibility, required knowledge, and available features.

At first, I also only understood it as 'something where you input text into an AI to create an image,' but as I played around with various things,

I started to think, 'Even with the same AI illustrations, the way they are made is completely different...'

So, what are the actual ways you can create AI illustrations?

This time, to make it easy for beginners to understand, I will introduce them by dividing them into the following three main methods.

This time, to make it easy for beginners to understand, I will introduce them by dividing them into the following three methods as follows.

1. Methods using web services and apps
2. Methods using conversational AI like ChatGPT
3. Methods for setting up an image generation environment on your own computer

Even though the goal of 'creating AI illustrations' is the same for each, the ease of starting, flexibility, and required knowledge differ significantly.

I would be happy if you could read this while thinking, 'Which method seems right for me?'

Now, let's look at them one by one!

2. 1. Methods using web services and apps (PixAI / Chichi-pui)

First, I will introduce methods for generating AI illustrations on websites and apps.

Web services that can create AI illustrations include, for example, 'PixAI' and 'Chichi-pui'.

I think this is probably one of the easiest methods to start with for those who want to try creating AI illustrations for the first time.

Basically, you just open a service that supports AI image generation, input the content of the image you want to create as text, or select from the provided settings.

There are many services where you can try generating images immediately without having to build a dedicated environment on your computer or make difficult settings.

Some of them can be used only with a smartphone, so

'I just want to try out what AI illustrations are like!'

it is also suitable for cases like that.

On the other hand, depending on the service, there may be limits on the number of times you can generate for free, or some features may require payment.

Also, since the available AI models and detailed settings are limited to what the service provides, there may be differences in flexibility compared to the method of generating on your own computer, which I will introduce later.

Even so, the ease of starting without any troublesome preparation is a major appeal.

If you are at the stage of wondering, 'What is an AI illustration like?', I think it's easy to start by trying out these simple methods first!

When you actually use a web service to generate, you can create illustrations like this!

This time, PixAI was used to generate it.
It's quite convenient to be able to create this much in a browser without having to build a difficult environment!

☆Seems suitable for these people!☆

・People making AI illustrations for the first time
・People who want to avoid difficult settings as much as possible
・People who want to try it out easily on a smartphone or regular computer
・People who want to try it for free first

3. Method ② Creating with conversational AI (Google Gemini / ChatGPT, etc.)

Next, I will introduce Google Gemini and ChatGPT, which are methods of creating illustrations using conversational AI.

Just like the web services and apps introduced earlier, you don't need to prepare a special generation environment on your own computer, and you can start relatively easily.

The biggest feature of this method is, after all, being able to create images while talking to the AI.

For example,

You can ask it to create an image, and while looking at the completed image,

You can ask for revisions just like having a normal conversation.

Even if you don't know much about the instructions called 'prompts' often used in image generation, you can proceed while conveying what you want to create in your own words, so I think it's an easy method for people making AI illustrations for the first time.

Also, even if you don't know how to give instructions, it's a convenient aspect unique to conversational AI that you can consult and think about what you want to create together.

On the other hand, you cannot adjust all the detailed settings yourself, and there are differences in the content, number of generations, and features available depending on the service you are using.

Furthermore, if you want to generate the same character multiple times or have specific details that you absolutely do not want to change, things may not turn out as you expected.

Even so,

“I don't understand the difficult parts yet, but I just want to try making illustrations with AI!”

I think this is a very easy way to get started for people who feel this way.

By the way, when I actually asked a conversational AI...

This was created by asking ChatGPT and having a conversation. The easy-to-understand part of this method is that you can make requests in plain language without knowing any technical settings.

☆Seems suitable for people like this!☆

・People making AI illustrations for the first time
・People with almost no knowledge of prompts
・People who want to create while consulting with AI
・People who want to modify generated images while conversing

4. (3) Method of generating on your own PC (Stable Diffusion / ComfyUI, etc.)

The last method I will introduce is preparing the environment for image generation on your own computer and creating AI illustrations locally.

Representative examples include "AUTOMATIC1111" and "ComfyUI," which can use Stable Diffusion-based models.

The big difference from the methods introduced so far is that you prepare what is necessary for image generation on your own computer and generate images in your own environment.

Because of that, you need to prepare software and models at the beginning, and a certain level of performance is required for your computer.

I am currently using ComfyUI myself, but when I first saw the screen...

“...Huh? What is this?”

That was my honest impression (laughs).

In ComfyUI, which I often use, you connect things called "nodes" to build the flow of image generation, but when you see it for the first time, you don't even know what to connect where.

VAE? KSampler?

Unknown terms keep popping up one after another...

However, as you gradually understand how to use it, you start to see the unique fun of this method.

Changing the model used (data that significantly affects the art style and expression of the illustration), adding LoRA (data for adding or adjusting specific characters, outfits, or expressions), and adjusting generation methods and fine settings are all part of it, and the fact that the range you can touch yourself is quite wide is a key feature.

...and here we go again with more unknown terms (laughs).

If I were to explain these in detail, I could probably write an entire article just on that, so for now, it's fine to just think of it as being able to 'customize things to your liking'!

Also, unlike methods that consume service points every time you generate, you process everything on your own computer, so another nice point is that you can try and fail without worrying too much about the number of images generated.

On the other hand, you need to learn how to set up the environment and how to use it, and if you don't have a computer suitable for image generation, the hurdle for getting started becomes high.

Therefore,

'I want to try making one AI illustration first!'

might be a slightly difficult method for someone to choose from the very beginning.

Conversely, while you are making AI illustrations, if you start thinking,

'I want to set things more precisely myself!'
'I want to try various models and LoRA!'
'I want to build the generation environment itself to my liking!'

then I think it would be fun to try local generation!

And this is something I generated using ComfyUI, which I usually use.

The initial preparation and settings are a bit tough, but in return, it becomes easier to finely adjust the parts where you think, 'I want it to be like this!'

☆Seems suitable for these people!☆

・People who want to adjust AI illustrations more precisely
・People who want to do various customizations with attention to detail
・People who enjoy the process of trial and error itself
・People who have a computer capable of image generation

5. What is the difference between each? Comparing ease, freedom, required knowledge, etc.

Up to this point, I have introduced three main ways to make AI illustrations.

'In the end, what is the difference between them?'

So, here is a rough comparison!

When you line them up like this, you can see that rather than one being the 'best,' each has different strengths.

If you just want to try it out easily, go with web services or apps.

If you want to create while consulting with the AI about what kind of image you want, use conversational AIs like ChatGPT.

If you want to be very particular and don't mind a bit of difficulty, go with local generation where you can tweak various settings yourself.

That's the general idea.

Of course, this is just a rough comparison. What you can do changes depending on the service or environment you use, and you don't have to choose just one.

I personally switch methods depending on what I want to create and my purpose.

I think the best way to start is to try the method that you feel you can handle!

6. In the end, which one is best for beginners?

I have introduced three methods so far, but

some of you might be thinking, 'So, which one should I use if I'm a beginner?'

Personally, if you have never made an AI illustration before, I recommend starting with web services, apps, or conversational AIs like ChatGPT.

Neither requires difficult environment setup, and you can easily experience what it's like to create images with AI.

Especially if you are in a state of

'Prompts? Settings? I don't really get it!'

then conversational AI is quite easy to understand because you can discuss what you want to create in plain language.

On the other hand, if from the start you want to

'adjust things in detail to your liking!'
'develop the same character further!'
'try out various generation methods!'

If that sounds like you, I think it's worth trying to set up a generation environment on your own computer.

There is a lot to learn at first, but in return, the range of settings you can control yourself is wide, and as you get used to it, you will be able to try various things.

I use various methods myself, but when you actually try them out, each has its own convenient aspects and difficult parts.

So, rather than looking for the 'best method' from the start,

I think it's fine to just try one that interests you first!

I think that's a good approach.

If you find parts where you think 'I want to do more of this!' while using it, it's not too late to try a different method then.

AI illustrations vary significantly depending on how they are made, so please try to find the method that suits you best!

7. Summary

In this article, I introduced the three main creation methods, starting from the question, 'How do you make AI illustrations?'

Even though we call it 'AI illustration' in one breath,

- Methods using web services or apps
- Methods using conversational AI like ChatGPT
- Methods using an image generation environment set up on your own computer

there are various options.

Each has differences in ease of starting, degree of freedom, and required knowledge, butthere is no single 'correct' answer that says 'you must use this!'

I think it's best to start by actually trying out the one that interests you the most or the one that seems easiest!

I wasn't knowledgeable about AI illustrations from the start either; I learned little by little while actually trying out various methods.

And even now, there are still many things I don't know or that don't go well (laughs).

In future Ruchimaru Notes, I might write articles that delve a bit deeper into each of the creation methods introduced here, or sharethe results of things I've actually tried and my failuresas well!

'I want to try making AI illustrations, but I don't know where to start...'

I would be happy if this article serves as even a little bit of a reference for those who feel that way.

Finally, I have lined up the images used in each chapter.
Actually, all of these illustrations were created using the same prompt (instructions), so please try comparing them once more!

Well then, that is all for this time!
Thank you to everyone who read this far!