AI Generated Images

You've probably been hearing a lot about AI lately.  It's pretty well unavoidable.  As such, there is more information about it that any single person can digest in a lifetime.  Still, i thought it would be a nice idea to tell some of my story and experience using it.  Because I've really come to enjoy it a lot, and perhaps people reading this can learn some stuff and try it out for themselves.

AI generated images, sometimes referred to as "Generative AI", is a means of providing a plain English prompt and the AI will go through its available image data and start combining things together that reasonably match what you asked for.  In experimenting with this, I created so many things that were just SO cool looking that I thought they were worth sharing.  I have an entire NFT collection now that is dedicated to images that were created using various forms of Generative AI.

You could go to your favorite AI generator right now and type in "person sitting at a desk writing a blog about generative AI, photorealistic".  Which I did.  And got this back.

Uh, okay.  Well, I didn't know what was going to happen.  But it's certainly good enough to make my point that it only takes a few seconds to get results.  And you don't have to know anything at all about how it works.

The funny part of this is that I have been creating graphics using Photoshop or something similar for like 30 years or something.  The process usually involves "I have this idea.  I need to find source images that look like this, this, this, and this."  Then I would get to work doing all the manipulation necessary to create an original piece of art from all of those sources.  Manual and time consuming, yes.  But that was always the best way to produce results.

I could have created a logo for a website I was working on.  Or I could have AI do it for me.  And so I did.

A few months ago, my boss came to me and said, "Can you create an image of a bald eagle standing on a mountain with the Philadelphia skyline in the background?"  So I did.  It took me all of 15 minutes, and I had a really nice graphic.  Then, it occurred to me.  That sounded precisely like the sort of prompt you would give a Generative AI.  And so I did.  The results that it came back with were pretty damned close to mine.  Only it took 15 seconds rather than 15 minutes.

This is a true story.  And I tell it because it puts a real world perspective on how Generative AI works.  It should give you a bit of a better understanding how these things do what they do.  It automates all that manual work.  It doesn't always mean that it will deliver what you want on the first try.  It doesn't mean that you won't have to go in and make some edits.  Also, different AIs are better and deliver better results.  Also, better prompts yield better results.  You can take classes on how to create prompts.  A buddy of mine does this and creates the most incredible pieces of art.

Enough of all that.  Let's talk about the different Generators used for the collection to which I referred.

Midjourney

Considered to be the premier image generator at this point.  They may or may not have a good web front-end.  I've only used it on Discord.  That was where this was created.

There is a reason that this is considered the best of the best.  It absolutely does produce the best results.  And they are always making improvements to it.  The only reason I don't use it more is that it costs too much.  Even the cheapest subscription plan is too expensive for me.  And that is because I'm not really making any money with this stuff.  I'm not earning anything to justify that expense.  In short, I need one I can use for free.

Fortunately, the others are free...  At least for the low-end. 

Blue Willow

This is also one that I've only used it on Discord.  That was where this was created.

This one works really well, even if it gives some absolutely surreal and bizarre results sometimes.  Other times, it produces stuff that is really cool looking.

Stable Diffusion

This one has a web front end.  And that is where this was done.

Another one that produces some really strange looking stuff.  And I find that the more intense the prompts the more insanity you get back.  I find that keeping it simple works best.

What else should we know?
Another thing that must be said about all of these is that AIs tend to have a hard time with certain human features like faces and hands.  Not to say that they are all bad.  Because sometimes the AI gets it pretty well perfect.  There are entire collections out there dedicated to 18+ naked people.  I won't link those up because this is a family site.  But you could easily find such things without my help. 

Anyway...  It's not unusual to see some truly horrifying human shapes.  And there is no telling when that will happen.  You just have to put in your prompt and hope for the best.  Animals, fruit, and mechanical objects seem to be among the easiest and most reliable prompts no matter what system you are using.

That's all I have to say about that for now.  I think this has gone on long enough.  I hope this has been helpful to you!  Now click on one of those links and get creating!

Here are a few helpful links if you would like to see more.

AI Generated collection on OpenSea

AI Generated collection on WAX

nft.durbnpoisn.com Marketplace

nft.dailynewspick.com Marketplace

ENJOY!!