Fixed :)
>>2953288p

@Background Pony #8255
Machine learning is used to train an AI on what are basically image tags.
It is provided an image, and all the tags relevant to the image are provided. This step is, at some point, human. Recently though, they’ve found out how to get the broad strokes of an images tags from it with a different machine learning model and then train that backwards into the image-generating machine learning model. It’s whacky. Basically though it is amalgamating tags to shape cues, color cues, and style guides. If it isn’t trained on a concept you provide it, it will “do its best” but its unlikely to be able to reproduce it. This is particularly the case for interactions between two characters in the same image. It also frequently has trouble with perspective, since turning tags into cues that get put into the final limitation cannot, broadly, generate a unique perspective. But it’s much more complicated than this. The important part is, rather than amalgamating images themselves, it’s amalgamating “learned concepts” of tags which come from people-made images. Usually.

They look kind of like a cheerleader outfit-straitjacket

I forgot the eyes >_<

@Background Pony #8255
Her arms are also her breasts. Because why not.

@Background Pony #8255
It’s literal magic

Whats up with her arms???

Really attempting to gather what this machines generated stuff are.
I don’t understand what i’m watching and how it’s made.

Now that’s what I call a handbra.