Most of my prior forays into creating images of real and imaginary cars using the Midjourney AI have come from Version 4 of the application. Version 5 is now in beta and there are definite improvements, from processing speed to much better rendering of hands. (Just do a Google image search for “Midjourney hands” to see the problem with V4).
An iconic car like the 1965 Mustang is something that that Midjourney V4 can render with a moderate amount of accuracy, but not without some issues. I’m no Mustang expert, but those symbols on the hood don’t spell FORD, and the horse on the grille just an odd blob of chrome. The six-legged horse does not approve.
Another V4 image has the same problems, and I suspect that those wheels are wrong. A rare horse/giraffe hybrid is keeping his distance.
Adding to the prior problems, V4 made the roofline a little too fast and added some 1966 bars to the scoops. The door handle seems to cover the door seam. The invisible horse next to the tree is unamused.
On to V5! The letters on the hood look like they spell FORD. The horse on the grille looks close enough to a proper one to me, and Midjourney seems to have learned the lesson, “don’t render a horse in a country scene if no one asked for it.” Those fog lights are far too inset, though.
V5 gave us a nice roofline. The scoops look good, as do the wheel covers. Well, the scoop on the driver’s side could be completely messed up and I wouldn’t know.
Another nice rendering by V5. I’m no Mustang expert so I’m sure the experts among you will point out what’s wrong, but this one would fool me.
The tires in the 3rd photo (horse in background) look especially skinny, almost like a VW Beetle.
I’d pay a few extra buck for some Tiger Paws
The last picture of the ’65 is VERY close. Even the wheel covers are perfect.
Looks like V5 may have ironed out a few bugs. Maybe you should go back and do the ’75 Packard Patrician again! 😉
Or Paul’s suggestion of the Square Bird in the next post.
Wait! – A 2023 or 2024 Thunderbird! 😃
It’s getting smarter, but it’s still got a ways to go. The green one (second to last) is missing its little rear side window; it’s got the wide C-pillar like a ’66 Thunderbird Landau. The last one is pretty convincing.
But I’m not sure what the point is; as for me it’s been all about the bizarre permutations its creative and deranged mind has come up with in the past. Presumably and hopefully it’s not lost that ability in trade of realism.
Stay tuned. I have something planned featuring what the AI thinks are either the ugliest or most beautiful cars in the world. Some of them are close approximations of actual cars, and some are… something else.
The readers task will be to distinguish which the AI deemed ugly and which it considered beautiful.
The entire exercise in using so called AI rendering seems completely pointless. From producing grotesque puddings to in this case photo realist facsimiles, the rationale is completely removed from the purpose of CC. The fact you can do something doesn’t mean you have to do it.
In the first picture the horsebug’s neck is impossibly Eschered with the fence slats. In the next picture the horsebug is balanced on top of one fence slat.