It is very bad. And it is IA , so it is worse. Since I don't like IA I don't think I could truly appreciate this gallery, but It is really bad in my opinion. And I am not sure if you make the pictures or just post them here.
In any case here it is an advice.
Make a story out of it.
Not like a comic, or a sequence, but let the picture be your story. You see, all of these pictures feel the same. They all talk the same way, they all say almost the same words. it is the same, there are exceptions, but the rest feels the same. It is boring.
So, which one is the scared one, which one is the defiant one, which one is the slutiest, which one is the shy one? and so, and so.
Look at the pictures and try to make a story from their faces, their poses, the backgrounds, the clothes, their expressions, etcetera.
The captions should reflect that story, and motivations and personalities. Not lore, or backgrounds and information, that is not necessary. But an unique take on the picture. A story, a personality, something that make readers feel like the girl in picture 1 is not the same as picture 2 or 3.
So they could talk in a different way, or use more complicated or poetic tones, or simpler and more direct talk.
What they say should reflect their situation. Were they sold? captured? is she convincing a friend or relative to join in? are they scared? is she challenging you? did she lose a bet? And so
Another thing is that they all use the same expression. I guess most of realistic IA training take their pictures from models, and models use almost the same expression in their faces, smile or serious, not surprise or fear or anger. That limits a lot of what you can do. The same with the poses, they repeat themselves over and over again.
That is my advice for using pictures from IA and caption.
bye

Base +6, Lady Luna Artemis +6, Fliphook +10, asdfgbvcgf -5, Rengaro +7, yupokamusi +8

I thought it would be something silly, but Monojungla's comment seemed like something very good and useful in the end. You should take note.

Base +10