Alright, so I found myself needing some images of cars from the 1920s the other day. Not just any old pictures, but ones that really felt authentic, you know? For a little personal project I’ve been tinkering with.
First thing I did, naturally, was hit up the usual image search spots online. Spent a good chunk of time scrolling through pages. Found some decent stuff, sure, but a lot of it was watermarked, low quality, or just didn’t quite capture that specific look I had in my head. You know how it is, sometimes you see it in your mind but can’t find the exact match. It was getting a bit frustrating.
Then I remembered all this fuss about AI image generation. Figured, why not give that a whirl? I’d messed around with it before for other things, sometimes good, sometimes, well, pretty weird results. So, I opened up one of those tools.
Getting Started with the AI
I started simple. Typed in something like “1920s car“. The results were… okay. Definitely cars, definitely looked old-ish. But they were kinda generic. Some looked more like 1930s models, others just felt a bit off, like a modern car trying to wear a vintage costume.
So, I realized I needed to be way more specific. That’s usually the trick with these things. I started adding details:
- Specific models: “Ford Model T 1925”, “Rolls-Royce Phantom I”, “Bugatti Type 35”.
- Colors: “Black Ford Model T”, “Deep red luxury car 1920s”.
- Settings: “1920s car driving on a dusty country road”, “Oldsmobile parked on a cobblestone street New York 1928”.
- Style: “Realistic photo”, “Vintage black and white photograph”, “Slightly faded color photo effect”.
This is where it got interesting. The AI started spitting out much better stuff. But it wasn’t perfect. Sometimes the wheels looked wonky, like they had too many spokes or were strangely shaped. Sometimes the backgrounds looked a bit modern if I wasn’t careful with the prompt. It took a fair bit of tweaking – changing a word here, adding a negative prompt there (like “no modern buildings“).
Finding the Good Stuff
After maybe an hour or so of this back-and-forth, refining the prompts, I started getting images that really hit the mark. Got a fantastic one of a dark green sedan parked by an old brick building, looked straight out of an old movie. Another good one was a black Model T kicking up some dust on a dirt road – that one felt really dynamic.
The key seemed to be layering the details. Not just the car, but the environment, the lighting (“afternoon sun”, “overcast day”), and the photographic style (“sepia tone”, “grainy photo”). It’s a bit like talking to a very literal-minded artist who needs exact instructions.
I saved a bunch of the best ones. Some needed tiny imaginary touch-ups in my head, but overall, I got a solid collection that fit what I needed for my project. It definitely beat endless searching for existing photos that weren’t quite right or had usage restrictions.
Final Thoughts
So, yeah. Using AI for this kind of thing? It works. Takes some patience, definitely. You gotta learn how to ‘talk’ to the machine. But it’s pretty cool that you can basically conjure up specific historical scenes like this. Way different from how we used to hunt for images back in the day. Saved me a lot of hassle this time around.
