These days you need something visual to add additional meaning, or credibility, or just whimsy to your blog posts. I’m relying on my own photos for personal remembrances and, more often than not, image AI creators when I don’t have my own photos; thus, no stock photos or licensed images. Here’s a growing list of explanations for how each came about.
About the Photos
Over the years I have horded boxes and boxes of photos, slides, CDs, and thumb drives of images I captured myself or treasures inherited from relatives who are now, sadly, no longer with us.
I definitely remember my mother wielding a Brownie-type camera (one of those dual-lens things, where you peer down into a finder). We had Kodak cameras, both one of those fiddly boxes where you had to carefully thread the film over a spool, and another where you just snapped a cartridge into place. There are the occasional Polaroids from family visitors.
From my early days in upstate New York, I have many lamentable photos and slides from my first 35mm camera, a Pentax K100. Starting around the dawn of the 21st Century I’ve owned two generations of Canon cameras which, together with a host of digital phone cameras, account for gigabytes of digital images. I’m plodding through them, revelation upon revelation.
In the early 90s I took a class in Photoshop at a local tech training center, learning just enough, as the cliche goes, to be dangerous when it comes to retouching. I tend to adjust levels and contrast and oversharpen; it’s just what I like.
About the Images
I’ve played briefly with Stable Diffusion, Dalle-2, Midjourney, and Microsoft’s Bing Image Creator. For now, I’ve settled on Midjourney. Though the Discord-based interface seems needlessly clumsy, I find it gives me something closer to my goal than the others.
I’ll admit I do not labor for hours crafting AI prompts, opting for “good enough.” Hopefully these aren’t laughably bad!
Now, On to the Credits
I don’t present myself as anyone with an eye for composition or a facility for restoration. I’m just a guy with some toys. I share here some background on any photos that needed rescuing or illustrations that sprang to life (yes, a strained metaphor) from what at the moment seems to me to be indistinguishable from magic: AI.
I was lucky that this was one of the first photos to see the light of day as I began spelunking through my boxes of memories. I have only the faintest wisp of recollection about the taking of this photo. I know that’s our front lawn, which is scruffier than it appears in my mind’s eye.
BEFORE. The original, faded photo was square. Being stubborn, I have as an ambition (obsession, really) that all my blog photos will be 3:2 aspect ratio.
AFTER. Photoshop 2023’s Generative Fill did a remarkable job of adding realistic details both left and right. Well, realistic enough. I spent maybe an hour fiddling with the fill and doing some color restoration and sharpening.
This essay presented the perfect opportunity for some tongue-in-cheek fun. I settled on Adobe Firefly.
BEFORE. This is the result of the simple text prompt, “Man with longing expression looking up into thought bubble that contains a bass fish.” I played with a few variations, but this seemed acceptable with very little work. Only problem? It’s square, and who needs that watermark?
AFTER. What better way to improve on the Adobe AI artwork than Photoshop’s Generative Fill. Only a few minutes to remove the watermark and produce my desired 3:2 aspect ratio, which also included the somewhat magical ability to build out the rest of the guy’s head. I know the ear is a bit wonky and his nose is a bit … unbalanced. (AIs seem to have trouble with appendages like arms, legs, ears, hands.) But I’m not expecting anyone to study this too carefully!
Using Adobe Firefly here. The prompt was approximately “Man no beard crossing river by hopping on stepping stones.” The AIs believe most men are bearded or at least not clean-shaven, and thus the reference to the beard. I didn’t take notes, but probably selected options for “Muted color” and “Golden hour.” It’s completely valid to point out that the guy here is one step from plunging into the water.
The original photo emerged from a box I’ve been moving, place to place, for decades. When was it taken? Probably early 70s. It is a shelter that was built on skids, enabling us to hook a chain to it and drag it with a tractor to another location. It’s not the same shelter where the action in this post takes place, but it’s built in the same style.
BEFORE. This is a portion of the original scan, cropped to the needed 3:2 aspect ratio.
AFTER. The major work was to open up the interior so you could see those gaps in between the side boards, a crucial detail in the story. The rest is just my usual updating to make things a bit more contrasty and a (bit too much) sharper.
This was the image that induced me to switch my AI generation from Adobe Firefly to Midjourney. To understand why, here’s the final prompt:
Closeup photograph of back of a young caucasian man’s head and shoulders. He is wearing a blue polo shirt. A woman’s hand is reaching down from the upper right, toward the man’s neck. You can see only the woman’s arm and hand. The woman has manicured pink fingernails. The woman’s index finger is inserted between the man’s neck and the collar of his shirt. She is pulling at his collar slightly. White background –ar 3:2 –v 6.0
Adobe Firefly kept responding that my prompt violated its guidelines. I suspect it had something to do with that bit about a woman’s finger inserted in the man’s collar. No matter how I tried to rephrase, Firefly evidently thought I was up to something inappropriate.
The prompt took quite a bit of crafting.
BEFORE. An early starting point. AI image generators famously have trouble rendering lifelike hands. And there was a lot of trial and error getting something that looked like a woman’s hand; adding the detail about red polish helped. Getting a finger disappearing between collar and neck, now that took luck and just respinning a chosen single image
AFTER. Even so, I had to give up on some details, such as the direction of the hand reaching down from above, which Carol Morris certainly would have been doing. Getting verisimilitude with an early 20s me wasn’t easy. I suspect my hair was longer and my shirt more rumpled. I finally had to stop here.