There's been a lot of chatter online these days about artificial intelligences creating original artworks based on text prompts. And I have been dissatisfied with my current artwork for the cover of Labyrinths & Liontaurs. I imagined writing a text prompt like "liontaurs walking through a labyrinth." So I went looking for an AI.
So I found my way to DALL-E And this is what I got.
a liontaur
![](lion103a.png)
a wemic
![](lion103b.png)
a lion-centaur
![](lion103c.png)
a centaur with the lower half as a lion
![](lion103d.png)
a centaur
![](lion103e.png)
a fantastical creature, like a centaur, human from the waist up, lion from the waist down
![](lion103f.png)
a fantastical creature, like a centaur, human from the waist up, lion from the waist down with four legs and two arms
![](lion103g.png)
a fantastical creature, like a centaur, human from the waist up, lion from the waist down with four legs
![](lion103h.png)
a fantastical creature, like a centaur, human from the waist up, lion from the waist down with four legs, lion paws, and a human face
![](lion103i.png)
a fantastical creature, like a centaur or sphinx, human from the waist up, lion from the waist down with four legs, paws, hands, and a human face
![](lion103j.png)
Of all of these, only one was even close to being a true wemic. But even this is flawed. What is going on with the back legs? Why no paws? Why is her head cut off? And I hate the facial features being so beastial. See:
![](lion103k.png)
DALL-E has an option to generate variations on an image, so I looked for variations on this one. None are better wemics. None are wemics at all, actually.
![](lion103l.png)
That's pretty much all I got to try, because DALL-E rations its free service. So bottom line -- at least this AI is not yet where it needs to be to illustrate my game! Maybe I'll try one of the others, or try again after the next update.