WWI nurse

I was so taken by Copilot’s beautiful rendering of a glamorous Black Elphaba that I decided to see if it could iterate a World War I nurse with the same “actor” — I thought maybe a uniform would boost its potential for a successful render.  

I was again pleasantly surprised with its output. 

Supernumeraries for an opera about the 1918 flu pandemic?

Apparently evenly spaced uniform buttons are a problem. 

As a piece of concept art this is amazing tho. 

Not what I was going for, but I admit I would joyfully watch this burlesque performance. 

It’s hit or miss whether it actually registers the physiognomy of a human figure in iterating tho. When I asked it to create an image of her dressed in a firefighter uniform, I got a bunch of white guys. Which I guess a lot of the extant images of firefighters it was trained on are white guys but that’s no excuse. 

It’s fascinating experimenting with this technology, and even though it is driven by conversational input, the output it creates does not follow in any path of human logic. You have to surrender so much creative control to the generative model, that I don’t know that it’s applicable to the creation of art where aesthetic details matter. 

Like, it’s a great way for someone with no confidence in their ability to draw to generate something that looks kind of like what they wanted, & it’s a great way to open up creative possibilities that would not have otherwise come to mind (like the WWI nurse burlesque concept art here). But I find myself continually frustrated by not being able to control something like the style of shoes she’s wearing, or the spacing of the buttons on his uniform.

I suppose a possible way to use it is to generate something that is close to what you want, then import that into a digital drawing tool like procreate and revise it yourself perhaps? I don’t know, this is an area of expertise that I have not worked in for something like a decade, so I’m sure it is something that designers of the future will figure out.