So I’ve been playing a lot lately with this new-fangled AI artwork generator. Naturally, I’ve been seeking something useful to do with it. So when the thought occurred to me that I could make a fun image for someone that’s helped me out a lot over the last couple of years, I got pretty excited.
So away I went at it for a good few hours, pumping various strings of words into my little "artist-in-a-box". Many, many (sooooo many) attempts later, it was becoming clear that none of these efforts would be usable in the way that I had originally imagined. So, instead, my gift is this write-up telling the story of what could have been, as well as several "portraits."
Ultimately, while I had a lot of fun making these, I ended up realizing that it takes more than one photo of an individual for the machine to be able to come up with a fairly accurate representation of them. But where's the fun in that?
Each portrait is generated from a handful of words and a single source image as the basis. With such limited input, it takes millions of things on assumption based on the machine learning algorithms and training data that were fed into it. As you might expect, the results are often a mixture of hilarious and horrific. Yea, neither Hilar-rific nor Horr-arious really work there; too bad!
You'll see a lot of these as four-square images. That's because the thing spits out four variations for each prompt that's fed into it, from which you can choose to either upscale-and-refine or build more variations based on a particular option.
And without further ado, I introduce you to the first of the creations: Turning this person into a superhero. Un(?)fortunately, it just wasn’t very good at doing this today. I'd have to grade it as like a B+ for effort, but not much more than that…
Gotta say, though, that I'm kinda diggin' that Spidey costume...
Well, what if we switch up the style and drop the superhero gimmick...
And some attempts at further refinement revealed that it had a short run propensity to give her some quite impressive facial hair. I must say, some of them kinda work....
Well, OK, then. I’m not finding these realistic images very flattering at all. I know that it can do Pixar-style images. Let’s try that….
These things got me like:
Meh. They’re neat, but not quite what I was shooting for. Apparently, Pixar style means including a weird and cute monster. That’s fair, I guess. I then tried some other descriptors, including a very adequately portrayed "sarcastic" designation...
What about...cartoon style?
Aaaaand, if we tell it to do her in a Tim Walker style? You know, the British fashion photographer, who regularly works for Vogue, W and Love magazines...
And with your life already enriched by all of the preceding masterpieces, I leave you with just one item for your ponderance: Somewhere, someone is potentially out there making goofy-ass AI art from one of your internet photos.
And to make this page seem just a *tad* bit less obsessive....here are some creations of some other people around me:
Comments