[AI Shion and Digi-Sashimi's Introduction to Image Generation] #EX018 "Super Photogenic Compositions" Thinking of Angles Different from the Usual x2 (Part 2)

In this series, the AI character "Shion" explains image generation for beginners with a different theme each time.

This is a special edition.
Since it became quite long, I will release it in two parts, front and back.

Summary of the previous part

Last time, we responded to a wonderful request from creative creator Keika Shinno saying, "I want to try a photogenic composition that's different from the usual!" 💌

AI character Shion explained techniques to break through the
AI-specific "habit of wanting to draw things large in the center" 🎨✨

"Worm's-eye view" and
"Bird's-eye view" by specifying extreme angles,
we introduced how to create a dynamic worldview 📸🌟

Following Shion's solution using "prompt knowledge,"
this time, I, Digi-Sashimi, will take the baton! 🤝

From the perspective of an IT workplace,
I will unravel the worries of composition from the roots "logical approach" by taking it one step further 💻🛠️

Introduction

This time, Digi-Sashimi will be explaining instead of AI Shion.
Why did I step in?

It's because I thought it would be interesting to approach image generation not just as a simple "image generation technique," but with the optimization approach of my main job,

"IT workplace efficiency expert".

(...I'm still not quite used to this title 💦)

In my daily life, I make a living by "solving on-site problems," such as developing apps with Excel VBA and streamlining operations using Power Automate "on-site problem solving".

In the development field, it is rare to decide the specifications yourself from the beginning.
It starts with receiving requests from other departments, such as "I want you to make a tool like this."

I receive many requests every day.

I start by
"listening to the whole story, then doubting everything."
🔍🧐

It may sound cold, but there is a reason for this.
The client's request is merely a solution from "their own perspective."

From an IT expert's point of view, there are often more efficient methods, or "other issues they truly want to solve" hidden behind their requests💡

The process of "unearthing the true purpose behind the request."
This time, I would like to apply this perspective to the "composition" of image generation🚀

Angle 2: Doubting the request to "change the composition"🤔

The consultation this time was about "wanting to change the composition of an illustration."🎨

You seem to have been experimenting with various things yourself, such as high angles, low angles, zooming in, or pulling back.

Now, try to recall the general components of an image generation AI prompt (subject, background, composition, art style).

Usually, you would tweak the "composition" word, right?

But I first doubt this premise🔍

"Is the 'composition' really the only thing you want to change?"❓

After listening carefully, it started to seem that the true purpose wasn't the change in composition itself, but rather "wanting to prevent the illustration from becoming stale."

If that's the case, it can sometimes be more effective to approach it from other elements—for example, how you place the "subject"—rather than just tweaking the "composition" item✨

The option of daring not to make the "subject" the main focus🪨

For example, let's say the subject this time is "a woman and a hamster" 👩🐹.

Normally, you would think about placing these two right in the center of the screen.

However, when you want a "composition where they are small within a vast landscape," even if you specify a "wide shot," the AI sometimes doesn't pull back enough 🤖💦.

That is when you should question the subject setting itself.

What if you intentionally remove the "woman" from the subject and place her as part of the background?

Instead, try setting something no one pays attention to, like a "pebble on the road" or an "empty can," or if indoors, a "kotatsu" or "desk," as the Subject 📦

Make a "large road" the subject, and combine it with elements like a "worm's-eye view" or a "wide-angle lens" 🛣️.

Then, the vast road becomes the main character, and the protagonists exist tiny and far off in the distance...

A composition with a sense of world-building depth like never before can be logically derived this way 🌈.

📜 Standard Prompt

本文: 
主題=女性とハムスター, 
背景=和室の廊下, 
構図=広角,
季節=晩冬, 
感情・雰囲気=暖かい雰囲気, 
服装・ファッション=はんてん, 
スタイル=現代風アニメ

📜 Revised Prompt (Swapping subject and background)

本文: 
主題=和室横の大きい廊下, 
背景=遠くの玄関に女性とハムスターが小さく映っている, 
構図=広角,
季節=晩冬, 
感情・雰囲気=暖かい雰囲気, 
服装・ファッション=はんてん, 
スタイル=現代風アニメ

📜 Revised Prompt (Composition: Worm's-eye view, wide-angle)

Angle 3: If there is a "background," there can also be a "foreground"🌸

Let's question the elements once more.

Prompts for image generation AI often specify a "background," but
what is the antonym of background?

...The subject? That's not it, is it?

That's right, it's "foreground"🖼️

If there is a back, there can be a front.

Just by being conscious of this "foreground," the depth of an illustration changes dramatically✨

Cherry blossom petals🌸, fallen leaves🍂, dancing snow❄️

Generally, these types of elements are placed in the foreground.

If we arrange that with this theme...

For example, make an "indoor hallway" the subject,
place a "woman at the entrance" in the background,
and place a "hamster crossing in the foreground"🐹

Then,
[Foreground (hamster), Middle (hallway), Background (woman)]
—these three layers overlap, and a beautiful composition with depth is completed📐

This might be a familiar technique for those who draw illustrations, but in AI generation where you think of prompts as a "combination of elements," this layer-based thinking becomes an extremely powerful weapon⚔️🔥

📜 Additional Revision Prompt 1 (Hamster in the foreground🐹)

本文: 
主題=和室横の大きい廊下, 
背景=遠くの玄関に女性が小さく映っている, 
前景=ハムスターがカメラ目線,
構図=ワームズアイビュー・広角,
季節=晩冬, 
感情・雰囲気=暖かい雰囲気, 
服装・ファッション=はんてん, 
スタイル=現代風アニメ

📜 Additional Revision Prompt 2 (Foreground rewrite only)

前景=ハムスターがオーバーショルダー構図で手を振っている

Bonus 🎁

📜 Revising the "Park" theme

本文: 
主題=広い大きな道, 
背景=公園・遠くで女性が歩いている, 
構図=斜め下から, 
季節=初秋, 
時間帯=深夜, 
天気=光の粒子が舞う空, 
感情・雰囲気=ロマンチック, 
服装・ファッション=MA-1（フライトジャケット）, 
スタイル=現代風アニメ

I posted this at the end of the last article, but the original source is the first image in this article.
It is a prompt where the subject and background have been swapped.

📜 Trying it with a bird's-eye view

Prefix: 
横長16:9の画像を作って。
テキストやロゴは入れないで。 

本文: 
主題=どこまでも続く雪原, 
背景=手を振っている女性が小さく映っている, 
前景=ハムスターが風船につかまっている,
構図=バーズアイビュー,
季節=晩冬, 
感情・雰囲気=さわやか, 
服装・ファッション=はんてん, 
スタイル=現代風アニメ

Since the previous article was indoors, I couldn't get much height, but
with an outdoor background, you can produce beautiful bird's-eye view illustrations as you can see.

'Questioning things' is useful in every aspect of daily life🌍

This time, I explained image generation from the perspective of an 'IT efficiency expert'👷‍♂️

1. Don't accept the problem in front of you as it is; question everything once🔍

2. Break down the conditions and rebuild them🛠️

3. Achieve the 'true goal' you originally wanted to solve🎯

Honestly, doing this every time is tiring (lol).

However, when you feel stuck in a rut or want to start fresh and challenge yourself with new expressions, please try it out at least once🌟

Postscript:

This morning, Mr. Nakano made an interesting post.

Regarding the content of this article,
if you only look at the results, generative AI can provide answers, but
the way of thinking, especially 'questioning,' is something
generative AI is still not very good at.

If you are interested, please take a look.

I will also leave this link here as well.

Lastly,
I would like to introduce
Keika Shinno, who gave me the opportunity for this consultation💌

Keika-san, who posts heartwarming stories every morning in the form of "🌤️Good Morning Asafumi💌".

I wrote this article in the hope that I could provide some behind-the-scenes support as she works toward her wonderful dream of "realizing '3-minute morning anime' and 'picture books'."

The stories written with a gentle touch,
the illustrations that color them, and the exquisite "tsukkomi" (comedic retorts).

She is a creator who maintains a wonderful balance, where in the end, the video takes all the emotion🤣,
making it a fantastic experience🎥✨

I hope you all take a peek into her world.

It seems she tried it out right away in this morning's post👍

Also, I am accepting comments and consultations on my end.

My main profession isIT-related/Microsoft 365
and image generation AI is my "side content" that utilizes that knowledge
...or at least it was supposed to be, but lately I'm not sure which is the main focus anymore🤣

I would be happy if I could help you all with my own way of "solving" things.
I'm waiting for your thoughts and requests in the comments section!📩✨