Message from @Eirshy
Discord ID: 628127190990454795
Nah it's literally a deepfake waifu generator hooked up to a money printer.
of all the fucking things an AI could/should be doing
I love free market...
fuck this, going to bed before I have another retard moment like last night and stay up past 5
Free market monetising boners
Okay quick question
how the fuck are you supposed to not sweat in a suit
it's fucking impossible
gonna be blunt I'm not impressed by the waifu generator. unless there's something deeper to it it's just a in depth created doll creator and those are a dime a dozen on the internet.
Name, and thirsty weebs that's how it got popular
I can make better waifus with the steven universe generator on doll divine.
also for the desperate furry there's a fursona maker jaesona put out. it's pretty good but has it's flaws.
@ryvergate Buy a lighter suit then
There are suits made for summer
I am literally wearing the lightest suit imaginable
back in the olden bronydays there was the pony generator. what it lacked in options for customization it made up for in poses.
The one I linked will at least have better art than anything calarts based.
Regardless it's more for novelty in that the artist is a neuralnet trained to match the style. The lack of discrete customization is because current AI tech does not understand meaning in the slightest.
This is also why it can't do animal ears worth a damn
And generally loses them entirely during posing
Honestly, the impressive part, tech wise, is the refinement process. Pallet swap is really simple and detail has vey high allowable drift, but the pose-setter is actually rather impressive.
I'd be more impressed if they bothered to explain how in a nuts and bolts way it's diffrent from your average doll creator in their blog.
Aren't the vast majority of "doll creator" apps just an extension of the sprite-and-transparency concept for 2D, and modular model/texture for 3D?
a lot of the small timer ones are but there are some rare ones that are a lot more complex than that.
it's why I sing the praises of the steven universe one. it has part and size scaling. the possibilities with that one are close to limitless.
This is literally just an AI trained to spit out quick eastern-style character portraits, followed by another to recolor, another to add detail, and a fourth to pose.
This isn't close to limitless, this is actually limitless as it is generating the images on the spot.
so those girls aren't canned?
And again, reason it lacks directly-requestable accessories: it has no clue what a "ribbon" is
Correct. The closest to "canned" you get is ~~the final result going into the trash where it belongs~~ an unknown level of pregen/caching being done to make the server scream less.
I would not be overly surprised if the very first step is actually just pulling n items from a list of a few hundred thousand.
WARNER BANS CLOWN MEDIA FROM JOKER DEBUT
> https://www.oneangrygamer.net/2019/09/warner-bans-clown-media-from-joker-debut/93949/
But the following steps are a little bit harder to automate without pregenning all possibilities for all images
I stand corrected then it's impressive on par with that su creator I keep talking about.
Er, not all but like, a few hundred to a thousand.
someday maybe all the best creators like this will mingle and we'll have a computer able to produce art on command.
Regardless shit gets ridiculous fast and is only 4-deep.
and then skynet will kills us all with waifus
Why would it?
Nah it will haves us pets with waifus to take care of us
Just be a fren to 2501... I mean SkyNet and you'll be fine