A.I. used to make digital art in Photography
Digital Photography and Post-Processing, 2024, Part Two
Can you stand another one? (All photos and enhancements by author.)
I am obsessed, I guess, with playing with the latest and greatest digital and artificially “intelligent” applications for photographers. I have never claimed to be especially creative. I don’t have much imagination. If I have a style, it’s bright, sharp, and colorful. I have no genre; I am that curious, oddball “photovore,” interested in nearly everything and, thanks to digital (unlike analog), able to photograph in any conditions. I am happy enough, but digital photography and amazing applications, augmented with artificial intelligence and machine learning, satisfy my otherwise diminished creativity.
Lately, it’s been Adobe Photoshop’s suite of AI Generative tools.
Reminder if you didn’t know: click on any picture to enlarge it.
I frequently pass beneath a small BNSF* (Burlington Northern Santa Fe) yard where sometimes a locomotive or few are parked freshly washed. I sneaked onto railroad property on a brilliant blue-and-gold day to photograph these. The pictures are superbly competent snapshots, especially after my standard processing, but jeeeeeepers, lookit all that junk cluttering up my scene, wires and poles and random stuff obscuring my beautiful engines.
*(Pronounced bin-sniff by aficionados.)
So I cleaned them up.
And as an afterthought, surrounded them with an artificially generated riot of bright wildflowers. The wires are gone, and clouds are added because I prefer them to a blank blue sky. Note, too; these engines, at seventy-something, are only a little younger than me, yet still thundering mile-long freights up and down the rails.
In my last piece on this subject, I discussed the ethics of doing this sort of trickery (skullduggery? Obfuscation?). I think I am an ethical person. I always try to disclose when I have “enhanced” an original scene. Miscreants who merrily publish deep-faked Taylor Swift porn are lower than scum, but they don’t care! They have zero ethics.
And I am still determining how they can be deterred, much less stopped.
I watched my granddaughter play at the Lincoln Children’s Zoo a few years ago. The zoo has dozens of friendly ride-on animals. I made this snapshot — pure unthinking grab shot — when another youngster popped in to ask if he could Ride the Tiger.
It was only a family snap in the family album of perhaps thousands of pictures of an only granddaughter. So, meh?
The other day, I came across it and wondered what to do with it.
Not even six months ago, Adobe Photoshop (Beta) introduced Generative Fill, with which you can do so many things it is boggling. It is in regular Photoshop 2024 now, at least the desktop version. You select your subject, then invert it to exclude everything you don’t want to see in the gen-filled render, then type in a prompt (in this case, it was the Indian veld), et voilà, the grandchild is yanking the ears of a dubious-looking but reluctantly cooperative tiger on the Indian subcontinent. Sort-of.
These composites are not one-click-wonders. There is plenty of experimentation, redos, dodging, and burning to present the illusion that the background is believable.
This great honkin’ varmint has been a friend of mine since I arrived in Lincoln in 2017. He’s a tad sad, but at least he’s not shoes and Gucci handbags. He’s also ancient, and the entire museum is closed for a long overdue remodel. My pal may be condemned to gathering dust in the sub-basement with all the other ancient, faded taxidermy. I will miss them all.
I thought to give him a new life in his comfort zone. Here he is, low-crawling ninjalike toward prey through a prehistoric mangrove swamp. My prompt asked for a reflection, e presto…
Good to meetcha; now I’m gonna eatcha!
Notice how the AI picked up on the ‘gator’s hindquarters, falling out of focus, and continued it in the background. The light on the critter was coming from the left; look at the light on the mangroves. Whoa.
All of this was doable, heck, even back at the beginnings of photography nearly 200 years ago. Still, it required skillsets I certainly don’t have, nor would the average photographer at any historical period. For me, the miracle is that I can do it now, by myself, when I never could before.
KEWL.
Other Amazing Advances in AI, especially like these from Topaz Labs
This is not a composite: it is a crop/enlargement/enhancement of one of the best-looking 20th-century hood ornaments, the 1936 Hupmobile Rocket Ship. I saw it in a gallery exhibit about the glorious Art Deco era. It was in a glass case so that I couldn't get closer. The entire frame is as you see on the left, but the picture I wanted is the one on the right, both horizontal and much tighter.
Sure, I could just crop and enlarge to whatever size and aspect I wanted, but the obvious limitation is the native resolution. I shoot half frames, 6000x4000 pixels, that will stand considerable crop and enlargement, certainly far better than the micro-sensor glorified point-and-shoots I started my digital journey on. Still, there are physical limits, just as there are in analog.
The picture below was cropped from the original, untouched RAW file. The noise (digital “grain”) should be evident on nearly any screen, especially if you enlarge it. This is why simply crop + enlarge is not the best practice.
So what is Best Practice? In analog, there really wasn’t much you could do if you could see the grain in a print from a normal viewing distance, defined as roughly double the diagonal of the print. We used to call that golf ball grain.
Digital has gifted us with many more means of reducing the grainy look, but there were limits even on the best post-processing apps — until A.I.
Maybe I spend too much time on my soapbox, but I am convinced that, used ethically and disclosed upfront, there is nothing wrong with using these fantastic tools. If you are in a competition forbidding altering a picture, don’t. If you aren’t caught, it doesn’t mean you did right. Photojournalists and evidence/forensic photographers are forbidden from altering anything, and there are severe consequences for doing so, so don’t.
At the Fort Peck (Montana) Interpretive Center is this full life-size Tyrannosaurus rex model that we visited in 2009. My camera then was another microsensor but more sophisticated, a Nikon P5000. I happily blazed away at the critter.
Clearly, I had a ball with this 2009 monster, all shot with a microsensor, then processed, denoised, and dropped into AI-imagined “prehistoric” scenes.
Apparently, I have a fondness for subjects with outsized serrated back-curved steak-knife teeth. This next big fella is Al, as all allosaurs get nicknamed Big Al. It just fits. They were a Jurassic very-distant ancestor of the much larger Cretaceous Tyrannosaurus rex. Allosaurs are quickly identified by the “horns” above and behind the eye sockets.
I found this specimen in (where else) a dinosaur museum and had some fun with him.
It took a while to get Photoshop Generative Fill to cooperate with my concept. Prompts with words like “bloody” and “mangled” caused the algorithm to seize up. The poor victim that it finally rendered appears to be a duckbill. One might wonder why a skeletal predator/scavenger needs to eat. Hmmm… (Maybe I do have imagination; twisted?)
And in conclusion… (oh, thank goodness)
I often mention that my transition from professional analog photography to digital started in 2007 with a $100 Nikon L12, a microsensor point-and-shoot that changed everything.
You can see from the date embedded (I eventually turned it off) that this was one of my early efforts. This is Sugar Loaf Bluff, nearly 85 feet into the sky on an autumn day in Winona, Minnesota. The only copy I can find had, at some point, been post-processed in Lightroom.
I did the 16x9 crop but also generated an infill at the bottom to add more reds and oranges. I gently enhanced the original sky to bring out the faint clouds.
📸As always, gratitude for looking in. I sincerely appreciate it! Questions in the comments will be answered promptly, so please ask.😊👍
