Dubious Quality

I wrote a few months ago about how I applied to be in the beta for Midjourney, an AI-powered art generation tool. 

I didn't hear back and forgot all about it. 

This morning, though, I received an invite to the beta. 

After a few hours, it's fair to say my mind is totally blown, and so is Eli 20.11s. It's incredible, and I'm not even using 1% o the capacity of this tool, because I know so little. 

Basically, you use keywords to generate art.

They could be simple, like "giraffe in desert," or incredibly complex, like "render of a futuristic Time Machine, blue fog ,dark, YELLOW LIGHTNING energy, epic, steampunk, metal, wood, presented nicely, rendered in octane, realistic, film grain, 35mm, 1600 ISO, Sony a7riv, 8k, unreal engine, dark arts, mythical, surreal. cinematic by Tsutomu Nihei ::Photoreal, Hyper real, ultra real, ambient occulsion, anti-aliasing, tessellation, volumetric lighting, dark, high contrast, sub-surface scattering."

I actually cut a few lines from that second description. It went on forever. 

The program will generate four versions of the image, and you can choose to upscale any of them (adding detail). You can also choose to generate up to four variations of one of the images. It's fascinating to watch people on the server refining an image in real-time. 

I thought of Gloria, who loved Frida Kahlo, and typed in "Charon ferrying souls across the river Styx in the style of Frida Kahlo." After a little upscaling, this was the result:

Interesting, but not great. Striking, though. 

This next one, though, is off the charts. Eli 20.11 put in "giraffe in the style of mad max," refined the result, and got this:

I love how the giraffe and the car are merged. Pasting the image into the blog makes it lose a little of its vibrancy, but the full size version is incredibly striking. Here's another one of Eli's:

What's most amazing about this is we both generated these images after barely even knowing how the program works. I've watched people who have a strong grasp of the program's potential generate stunning images. I honestly would not be able to distinguish them from "real" art. 

This is incredibly disruptive technology, and I'm going to spend way more time than I should working with it over the next week. Eli, too.