Some might remember when I played with Deep Dream Generator a few years ago. If you need a refresher, read THIS post, and, if you are a masochist, links to all related posts are HERE. The SmugMug Gallery collecting all those images is at this LINK. Warning: there are 382 images, and while many are (in my opinion) very good, there are many where I was trying out stuff and learning.
What is Deep Dream Generator? Well, it’s a bit like what I do with Topaz Studio when I artsify my images. Essentially, it blends the content, style, and colors of one image with the style of another to generate a mishmash with elements of both.
For the sake of this post, here are a few images to give you a flavor of what Deep Dream could generate:
Some of the photos are small because you ‘earned’ abilities and options the more you used the app. By the way, the site is still active; it’s just that I lost interest in it and moved on; however, my account is still active (HERE).
Well, a few weeks ago, I came across THIS article. It sounded interesting, but I didn’t read it until last night . . . and, after reading it, I had to give a few of those a try. There are two I’m still waiting on (it’s by invitation), but I was able to jump onto and use three of the mentioned apps.
For them who don’t click on links, rather than blending photos, you tell the AI what you are interested in. Meaning, you use words to generate art; you write something describing what you’re imagining, and the AI renders an image.
The first was NightCafe AI Art Generator. It’s OK, but it works on credits (you earn 5 to 13 credits depending on how you sign on, and then you use those credits to generate stuff; when you run out of credits, you can buy more). This was the first thing I tried, and I think I used the words “Firefly Serenity ship and Hummingbird” or something like that, and this is what it generated.
It’s not particularly impressive, and it’s small . . . and it’s low-res. I created five renderings, and I’ll show them below, but I’ll actually showing a modified version of what I saved, enlarged by a factor of two (using Topaz GigaPixel) and cleaned up with Topaz Denoise AI. Here they are (including the larger version of the first one):
Frankly, I was disappointed with this one, and I won’t be buying any credits unless they improve substantially. Click HERE for a slideshow of the NightCafe AI Art Generator Gallery; it has the original-size renderings, and then the same renderings as enlarged and cleaned up by me.
Edited to Add: each day, you can claim five additional Credits by going to the NightCafe site and clicking “claim credits”. So, I’ll have a few more chances to try something more involved and see if they get any better.
Next up, we have Dream by WOMBO. It’s completely free (for now), and again, you use words to generate graphics. In this case, they generate trading cards. Unlike NightCafe, these are a good size and interesting, although they don’t match the originating text all that much (apparently, the AIs don’t know nothing about Serenity, the ship from Firefly). Still, very interesting. There are eight cards.
They might not be to everyone’s liking, but I was pleased with all these. There are many styles I’ve yet to try, and I look forward to playing a bit more with this AI. The SmugMug Gallery is HERE and if you want to see a slideshow of those cards, click HERE.
You know how people say they’ve saved the best for last? They’re not kidding. If you want to see an interactive explanation and examples of what MidJourney does, watch THIS VIDEO. Or, you can do a search for it and get various articles and stuff.
I’ll be honest . . . there’s a bit of a learning curve to getting to use the app, and at times, it was a tad frustrating. Mind you, it’s not the tool itself. It’s joining the community and then figuring out how to run the app (there’s a lot of help online, but still, not as easy as the others . . . or, maybe, I’m getting old and losing my faculties). They just opened up to “public beta”, so you can join and try it out and generate up to 25 renderings before you need to subscribe.
Briefly, you type something like “Firefly Serenity Spaceship over Pikes peak with a swarm of hummingbirds in the nightsky“, and the AI gives a choice of four renderings.
These are rough renderings and small. You can either ask for a different four or ‘upscale’ one or all four. I upscaled the first and third . . . and this is what it rendered (I suggest clicking on the photos for a larger view) . . .
OK, I was pretty pleased with those . . . so I tried this prompt: “Michelangelo drawing of Serenity Spaceship with Sarurn in the background” . . . yes, I misspelled “Saturn” (hit the wrong key), but when I got these four choices, I didn’t mind. I even didn’t mind that the AI had no idea what the Serenity Spaceship looks like . . .
I picked the first, third, and fourth for upscaling. I should mention, you can upscale one or all of the choices, but you then also have the option to upscale that result to the Max. The first two I showed, as well as the next one, were upscaled to the Max. Each upscaling counts toward one of your 25.
Excuse my language, but . . . Holy Crap on a Cracker! It’s as if Michelangelo himself is hiding somewhere and quickly generating these.
As mentioned, that’s upscale to the max. For the second choice (#3 in the group of four), I’m showing both the regular and max upscale. They are similar but with different details.
Really, either version would be good enough, but I was just Max Upscaling everything.
I mentioned that you can also choose to do a variant on any of the four, and I did that for the first one. When you do that, you get another group of four based on the one you picked.
. . . but let me show you the other one before I do the variant.
Here are the variants on the first . . .
I chose Max Upscale for the first one . . .
All of these are as generated. The gallery has duplicates that have been run through DeNoise AI (I’ll have a slideshow at the end).
The last one I tried before setting down to write this post (I wanted to share these) was the following prompt: “Michelangelo drawing of mechanical hummingbird, realistic, detailed.”
I picked the first three to upscale . . .
Needless to say, I bought a subscription. I don’t know how long I’ll keep it, but I can see myself
wasting time enjoying playing with MidJourney.
As mentioned, the SmugMug Gallery (HERE) has a duplicate of each rendering that has been run through DeNoise AI. The difference is subtle, but some are improved (in my opinion). You can be sure I’ll play with additional post-processing to see what I can do with these.
For now, here’s the MidJournal AI Art Generator Gallery slideshow.
That’s it. This post has ended . . . except for the stuff below.
Note: if you are not reading this blog post at DisperserTracks.com, know that it’s copied without permission, and likely is being used by someone with nefarious intentions, like attracting you to a malware-infested website. Could be they also torture small mammals.
Note 2: it’s perfectly OK to share a link that points back here.