I have recently been experimenting with a new technology, I know, I'm a techy geek and your are not shocked that I'm geeking out on something new. There are many new technologies harnessing AI to generate images based on words and instructions you feed it like Dall-E, Dall-E Mini aka CrAIon, and Midjourney. I got on the beta of Midjourney.
Midjourney runs using Discord (desktop or mobile) as its interface using the bot to send the commands to their trained AI engione. Its fun to learn and just watch and see what others are creating and how you can string commands to a specific look or rendering engine to generate.
How you use Midjourney
You use the command /Imagine <then you place a description what you want to imagine and get a result of>
For example I used:
”bioluminescent lupine flowers and morel mushrooms among a deep old growth ponderosa pine forest with dramatic lighting”
You get 4 images returned in a matter of a minute or two. You then can select and upscale the one you like, or select some iterations of some of the initial returns.
I will select the 2nd and 4th to upscale to high resolution using U2 and U4 as well select V2 and V4 to get more iterations to see more iterations. And thots how you lose hours of time geeking out.
Its quite interesting, and certainly not perfect, especially with faces, but can be a fun experimentation of how an AI interprets the text statements based on training data on the internet. You can have a dog made of marshmallows dancing on a river of chocolate and gumdrops or nautilus inspired iridescent buildings. I personally don't think this is taking any artists or designer jobs but may someday be an AI assistant to help concepts or perhaps study even building or design ideas but that is a long ways to go. AI can be used in many ways to help artists and designs such as mocap assistance or cleaning up images, or helping you in-context with something you are designing and offering options or looking ahead to what you may need next or checking standards. But for now this is pure geek fun to play with and see what weird results the AI comes up with as it can seem like someone spiked your coffee or you ate the wrong mushrooms in the woods. But, do end up with deep questions like “how does the AI know that a hat goes on a head, even a fish head and not the tail”. More bizarre is asking open ended like concepts like “love”, “peacefulness” or “the day the universe ended” and see how bizarre and thought provoking the result and try to think how and why it came up with the results.
I asked it to imagine an illustration of a silver labrador dog and got this one eyed creation.Still better than I can draw, especially in 2 minutes, but I at least know a dog has better eye placement than this zombie mutt.
Or my skyscraper made of cats and fish:
San Francisco in the year 3000
I would not recommend asking “average people of your state or country” as the AI is rather weird with the results. Perhaps because there are so many weird internet photos out there or the AI just thinks poorly of us humans.
Here are some of my creations the past 24 hours of playing around. Some are cool, and some are spooky:
Sign Up for the Midjourney Beta