Learning the Basics of YAML. A Tutorial with Practical Examples 🎨

This is a character I made recently, who will be appearing in this article ( 'ω' ).
I usually generate non-human characters, so it's been a while since I generated a proper human character.

I did design her with serious intent, but I'd give the result about a 30 out of 100.

In other words, anyone can create output that is ‘exactly as intended down to the details’
like this.
Since it's AI, there will be some variation, so it won't be exactly the same, but it should be roughly around this level.

Preparing the materials 🔨

Up to this point, I've been writing only in Markdown.
From the draft to the final version, it's not that difficult even if you write it by hand.
It's better to write it yourself until you can imagine it in your head.

I was the same way, but you learn better that way.
Once you get used to it, you can leave it to the AI, but this experience will also be useful when checking the quality of the results.

First, please fight with it.
Make plenty of mistakes, too.
But make sure to save those failed images and prompts.
They are all fragments of the knowledge you have built up.

By looking back at them later or comparing the differences,
you might discover new insights or techniques.
You remember things better through practice than by hearing about them from others, right?

As an aside,
even a ‘collection of failed images’
is something that people on social media actually enjoy looking at.
It might actually be easier for them to empathize with that ( 'ω' ).

Trying to imagine focusing on the details 🌈

I also described the seven colors of the sword in YAML, but the structure is simple.
I just followed the design principles for prompts, organized things logically, and defined the framework and flow.

What is this person talking about?
I have no idea 🌿

To you, who thought that,
your reaction is normal ( *´ω`* ).
If it were me last year, I would have had the same impression.
To get this far, I've been fighting with the AI constantly.
It's a fight festival. And every day, endlessly 💦

What I was particular about was the seven-colored effect ⚔🌈

By making the background a single color,
the AI only needs to allocate a small amount of resources to drawing that part.
When you want to make the most of a specific detail you absolutely don't want to compromise on,it's also a valid strategy to kill off the other parts, like in this background example.

However, with current GPT-based image generation AIs,
previously, the issue of 'inconsistency when adding more characters' has been significantly improved.
As shown in this image, it has become easy to reflect many characters.

The seven-colored effects in this image have also become more high-precision, and
I'm glad the parts I was particular about turned out well ( *´ω`* )
That is, after deciding that 'this is within the acceptable range,' I maximized the output for this effect part.

Summary🌱

As with any AI utilization, the premise is not to expect a finished product on the first try.
The correct approach is to adjust and bring it closer to completion.

Next time, I will finally explain text generation with practical prompts.
This will be a course on how to write in a way that is 'better understood by AI' than even Markdown🤖
Learning this will also lead to an understanding of the concept of the simplest hybrid prompts.

Thank you for reading.
If you liked it, I would be happy if you could like and follow me🐈‍⬛
I sometimes announce article updates on X (｡･ω･)ﾉ
If you follow me there as well, I will be overjoyed✨💕