I wanted something for my December calendars, and the prompt was simple: winter oil painting owls. Above, is the first of four images MidJourney Version 4 generated. We’ll get to the others in a moment, but first, a bit about how I use the tool.
I process the initial version (above) through Topaz GigaPixel, doubling the resolution, removing some noise, and increasing the sharpness. This, then, is the result of that manipulation:
Unless you click on it, you wouldn’t know it’s twice as big as the first version, but you should notice it’s “cleaner and sharper” . . . unless you’re reading this on a phone or other small screen. In that case, the difference probably isn’t noticeable.
By the way, I linked large images . . . they might load slowly unless you have decent Internet service. It probably won’t help that this is a long post. If you want only the images, scroll down to the bottom and run the slideshow.
You might wonder why I bother with using GigaPixel to increase the size and improve the image . . . it’s because I like the initial rendering, and I know that the MidJourny upscaled version will not be the same. Here’s the MidJourney upscaled version of the first image:
Perhaps that’s closer to what an oil painting might look like, I don’t know, but the features are not as smooth. Mind you, it’s still an impressive image, but not — to my eyes — the same as the original. By enlarging and cleaning up the original, I now have two large images, both of which I like.
Once upscaled by MidJourney, I typically have it ‘remaster’ the image. In this case, the remastering gave me this:
I wanted to show this because it’s a lead-in to discussing these tools.
Some readers might remember THIS post . . . That was in August, and as poor as those renderings were, the promise of better stuff was there. Look at THIS SmugMug gallery Slideshow (17 images) of a similar prompt recently done on my phone.
Now, if you’re an illustrator, this tool might give you pause and perhaps have you consider changing careers. Fair enough. At the very least, it’s competition, and at worse, you might think it will supplant you.
Certainly, for casual customers, this might be a better avenue than hiring someone . . . but, let’s face it; casual customers were never going to hire an illustrator. At best, they might look for images or renderings in stock photography and stock illustration sites.
Well, what about those sites? Are they in trouble? It depends. Casual users typically look for free stuff. On the other hand, content providers license images, music, and any media they need to produce what they want to sell. Content providers aren’t selling media; they sell a package that might include some media as a visual aid.
Someone needing media for what they do isn’t all of a sudden going to take it upon themselves to create the media; it’s usually not cost-effective.
I mean, some might, but even this powerful tool requires time to master and use effectively, and you don’t always get exactly what you need. Heck, sometimes you have no idea what you’ll get.
By the way, here’s the second rendering of the four I mentioned . . .
Again, amazing stuff . . . but I didn’t control the number of owls, what they look like, the color, the setting, etc. And look at the change between the original and the upscale, and then the remastering. Vast differences.
All fine and good because I’m playing around and not using it for anything other than posting here, but what if I had a vision for a magazine article or product I wanted to sell?
OK, OK, for many, it might not be different than buying a stock illustration or photo, you see something that might fit, and you use it.
But again, it’s likely that the same people who currently sell photos (photographers) and illustrations (graphic artists) are likely to sell these creations as well. They have connections, established reputations, methods to deal with licensing, etc., etc.
In fact, many people using these tools are already setting up shop and selling prints and digital files of their creations . . . but they pay more than what I do to create these offerings, and they lock them down with private prompts and copyrights . . . just like you would for people-generated illustrations and photos.
I don’t see these tools changing much of the commercial aspects of stock media. People who make a living from selling media will probably incorporate this tool into their arsenal of other tools.
Wait . . . what about users like me?
Before I answer that, let me show you the third of the four owl offerings . . .
Well, I’m playing with the tool and having fun. Mostly because it’s new and partly because it’s constantly improving, and I want to see what it will create.
It’s a bit like fishing . . . you cast your line out there, and you don’t know what you’ll pull up. Maybe a tasty trout, or maybe an old boot. Perhaps it’s also a bit like gambling; you roll the dice and see what you get. The thing is, while I dabbled in both fishing and gambling, I quickly lost interest in both, and I no longer do either.
“These are great! You should try selling them!”
That’s the same refrain I hear about my writing, my photo, my mind, my body . . . oh, alright, it’s only the first two, but the reasons I won’t market this stuff are the same as the reasons why I don’t sell photos . . . this is pretty simple stuff, and I’m not serious about it.
There are many, many, many better users of these tools, just as there are many, many, many better photographers than me out there.
But, say I wanted to sell these . . . first, I’d have to upgrade my account at the sites ($$), then set up a site to sell my creations ($$), then advertise ($$), then hope customers come flocking . . . hmm . . . dealing with customers . . . better I hire a lawyer ($$) to help me write a contract of some kind. And, all the while, I’d be competing with people who do this for a living and have been doing it for much longer than me, have connections in the industry, and are more driven than I will ever be.
Before I give my take, here’s the last of the four renderings.
The gallery has an enlargement of that last image, making it huge.
Anyway, my answer is that AI Art tools are to photography what frozen dinners are to cooking. Some will indulge, and given the current trajectory, I can see where renderings might be as good as photographs, but it won’t matter.
Computer graphics already offer amazing capabilities, but people still paint and draw. Same here.
Maybe not for all, but for many, there will be no substitute for the tactile and emotional connection between the photographer, camera, and subject. That connection is what makes me like my photo of a cardinal more than a photo of a cardinal by a professional photographer, despite the professional photographer’s photo being the better photo. By the way, I mean the bird, not the church official.
Same with these renderings. There are literally tens of thousands, if not hundreds of thousands, of these renderings out there, and some are truly fantastic . . . but I’m not out there looking at them. I have zero interest in them . . . because they are not mine.
Well, not exactly zero interest . . . like with other people’s photos, they can be an inspiration for what can be done.
I’ll start writing about my Panama trip soon, and my desire to do so is partly driven by my desire to share my impression and photos. If all I wanted to do was to have readers see what the Panama Canal looks like, I could point them to any sites (or Wikipedia) and not even bother writing about my experiences . . . but those sites won’t reflect what I saw or how I saw the Canal.
Will these tools have an impact on how we do things? Probably. After all, I used a few of them in my December calendars instead of strictly using my photos.
But why did I do that? Well, because the idea was to share calendars, and these renderings helped me do that. And, these creations were still the result of my effort instead of me grabbing them from elsewhere.
Would photos have been better? Maybe. Maybe not. The calendars look OK, and that’s all that mattered for that post.
By the way, one of the images I used for the calendars was based on this prompt . . . “winter oil painting ice skating on a lake”.
I got four renderings and none with what I considered a depiction of ice skating. I think the tool still has issues rendering realistic human or animal figures in a scene. They can do amazing portraits, but people in a scene is still an iffy proposition . . . but I got an image I liked.
For the above, the remastered version is my favorite . . . but then, I typically don’t have people in my photos, so that’s no surprise. It’s worth repeating, the tools seem a long way from being able to show realistic crowds or even small groups of people in a scene. I suspect they might eventually get there.
I’ll be playing around with all the models the tool offers, but the point is that while I like this rendering, it will not supplant my desire to snap photos.
“Yeah, but do readers prefer this rendering or your photos?”
It depends . . . perhaps they like both, since they are both ‘mine’. Perhaps they will like the rendering more than they like my photos. Perhaps they will think both are crap and don’t even follow this blog.
I post things that interest me, usually things that I like. Judging by the traffic to the blog, about ten other people find what I publish interesting. I don’t know why, really. Perhaps they’re just being polite.
Some readers might stop by because of the content, and some might stop by because they “know” me. For some, it’s a combination of the two.
Will I be supplanted by some AI-generated amazing blog posts? Will readers grow disinterested in my bland offerings? Perhaps, but given there’s already amazing blog content available out there, and I’m still publishing posts and getting a few views, I don’t think the addition of AI-generated blog posts will affect what I do.
Let me share a few other things, thus making this post rival some of my mega posts of yesteryear (not really — it’s not even 40 photos).
I mentioned people still paint, sculpt, and create things by hand. For example, Melisa quilts, my mother does cross stitching, my sister Louise likes to scrapbook, and my sister AnnMarie likes to watercolor.
Let’s take watercolor . . . I gave the AI this prompt: ornamental watercolor winter scene cabin mountain lake.
I got back four renderings. Here’s one, along with my edits, the upscaling, and the remastering (click images for larger views to see the difference).
I’m no judge of watercolors, but I can say those look nice (the 2x version of the first is my favorite).
But, say AnnMarie painted that scene . . . well, it would then add another dimension to the drawing; a personal connection.
Now, say someone comes to me and says, “I can show you a similar scene painted by the foremost watercolor artist in the world!“
My answer would be, “That’s nice, but no thanks.” The reason for that answer is twofold; I have only a passing interest in watercolor art, and I have no familiarity with any famous watercolor artists. Meaning, I have no connection to their art.
If I were a watercolor artist, I would feel differently. I would be interested in learning about what masters of the art are doing.
For instance, I follow some blogs because they offer photos that are different than mine. I get ideas from them. I learn techniques from them, and I see what’s possible. I can appreciate what it took to generate the photo because I’m familiar with the process.
For a moment, let’s compare AIs to Ansel Adams. I can look at Adams’ photos and admire them. I also immediately know I can’t compete with his work.
But, even knowing I can’t compete with Adams, I’m not hanging up my camera . . . because it was never a competition in the first place.
Let’s look at another AI watercolor rendering (the last one, I promise) . . .
A little sharper, a little cleaner . . . it’s the version I like best because I’m not an expert in watercolors; I just know what I like.
Here’s the upscaling by MidJourney . . .
This is also nice, but noisier, rougher. And here’s the remastering by MidJourney.
Again, not as well defined and rough details with more noise . . . so I cleaned it up a bit . . .
Which of those two is better? For me, the second one, but the “best overall” is still my initial enlargement.
I recently showed a series of different processing for a scene with some leaves changing colors.
Different people liked different renderings, but a few people mentioned they preferred the natural scene more than the renderings with more processing. The thing is, I offered up a choice . . . had I shown only one version, people’s reaction to it would be different than when presented among other choices.
It’s probably the same with AI-generated art. On its own, people will feel one way about it, but when shown next to people-generated art, they might feel differently. Some might prefer one or the other. Some might appreciate both. Some might not care for either.
Really, it’s no different than it’s been since the beginning of artistic expression.
To this day, I don’t understand what people see in Picasso’s works . . . but I know he would have been a failure as a police sketch artist.
Here’s the slideshow of the gallery with the above renderings (and a few more):
If you want to manually scroll through the gallery, click HERE. Hover the mouse on the right or left side of the image to show the arrows to scroll through the images.
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.