[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/ic/ - Artwork/Critique

Search:


View post   

>> No.6414914 [View]
File: 230 KB, 1468x920, samdoesartsAIbash.jpg [View same] [iqdb] [saucenao] [google]
6414914

>>6414911
And lmao :))

>> No.6413905 [View]
File: 230 KB, 1468x920, samdoesartsAIbash.jpg [View same] [iqdb] [saucenao] [google]
6413905

>>6413882
>it's not even the same face
>same beard, same face, head inclined at the same angle, looks 99% like the image in question
>i-it's not the same broh, it's just not, look, the pixels aren't even identical IT'S NOT THE SAME OKAY?!?!?!?!
And the denial continues.
I got it from here, by the way- https://www.artstation.com/artwork/14zy3K
Coincidentally, it was one of you fAIggots shilling in a thread that lead me to this, so you have my thanks for that, whoever it was.
By the way, while we're here, let's just pull out another clear case of SD generating what is undeniably something that it was trained on. I also got this from an AI fag that posted it as an example of the SamDoesArts dataset. I don't have to do this myself, you guys just shoot yourselves in the foot for me.

>> No.6401032 [View]
File: 230 KB, 1468x920, samdoesartsAIbash.jpg [View same] [iqdb] [saucenao] [google]
6401032

>>6400677
>>6400527

Oh really? Then explain this, smartass. I have found two instances where it's done EXACTLY that, and I wasn't even trying very hard (and in picrelated, I found this shit by ACCIDENT because it used a meme image that almost everybody knows about). It's not literally photoshopping fullsize images from a multi-terabyte database of images, no, we understand that you fucking pedant, but it's definitely recording these images in a highly compressed way and reusing them wholesale. It's not creating anything new, it is, for lack of any better description, photobashing. So don't UM ACKSHUALLY me, faggot.
EDIT: Nevermind, someone already uploaded it before me. See this image here for "Sad Keanu Reeves as King Arthur wounded and bleeding on a field.jpg" Looks like it's getting around, good. So I'll just use the SamDoesArts one instead.

>> No.6395975 [View]
File: 230 KB, 1468x920, samdoesartsAIbash.jpg [View same] [iqdb] [saucenao] [google]
6395975

I CAN'T BELIEVE THAT THIS IS WHAT WE'RE AFRAID OF REPLACING US
OH AM I LAFFIN'

>> No.6385282 [View]
File: 230 KB, 1468x920, samdoesartsAIbash.jpg [View same] [iqdb] [saucenao] [google]
6385282

The thing is that the Stable Diffusion AI may be a lot more derivative than one might think at first glance. The more limited the data it has to pull from- such as from a single artist, or a niche style that doesn't easily match up with real life photographic data like ye olde master realism does- the more you can see the "seams" and instances of almost blatant copy-pasting where it lacks material to work with. See this comparison of a SamDoesArts piece with something generated by a SamDoesArts SD model, which I found within about five minutes of searching- it straight up looks like a shitty photobash and paintover.
So no, this isn't about "muh style". This thing is straight up stealing your shit and photobashing with it.

Navigation
View posts[+24][+48][+96]