I would describe these as mixed results. Sometimes my little AI's can do amazing things, but sometimes they are not entirely up to the task. And sometimes, to my bitter disappointment, they can't do it at all. For instance I spent hours trying to get a t-shirt graphic of a person, eventually my cartoon character, craning his head around to take a bite of a taco, while the carefully assembled taco was held perfect and horizontal and undisturbed at the center of the image. And the title would say: "The taco does not come to you. You go to the taco."
Was this idea so appealing that I would spend hours trying to get an image of a person craning their head around to eat a taco?
Perhaps sadly, yes.
Did I succeed?
Certainly sadly, no.
I could ask this AI to write a poigniant novella in Russian about an obscure blogger and it could probably do it. But someone eating a taco? No.
This idea today I have been working on for a couple of weeks. I call them sketchbook pictures. I'm not totally satisfied, or maybe I haven't gotten to just the right application of it yet. But unlike the person eating a taco pictures, I at least have some things serviceable enough to show you, and plenty of them.
