AI Redraws Boston

That first picture is charming, and also not completely detached from reality. Makes me think "looking north up the Fort Point channel if the USPS facility were redeveloped in the style of the Leather District".
 
kdmc said:
That first picture is charming, and also not completely detached from reality. Makes me think "looking north up the Fort Point channel if the USPS facility were redeveloped in the style of the Leather District".
Click to expand...
Definitely some kind of alternative reality going on there, LOL.
 
The second one is really messed up after close examination. The physics of the flags to the far left make no sense. Check out the "American" one that's really two, except also they seem to be sharing flagpoles with other flags. Then the sign on the street says "BOSTN" and the rest of it is in another language.

Then there's the race itself, with all the fans in the middle of the street and people seemingly competing on either side of that metal gantry. It seems like people on the other side of it are actually running toward it, and everybody is going to collide in the middle. Are there more competitors behind all the fans crowding the foreground? Where is security? Why does the guy squatting at the "BOSTN" sign look like he peed in the road?

The city itself looks cool though. This AI understood the "Athens of America" look.
 
That second one reminds me of a 1930s/40s propaganda poster for the Third Reich. The heroic grandiose look is over the top.
 
Here's a different program.

1704511392775.jpeg

 
I had played around with Bing's image generators a few months back. It came up with some cool stuff for "IndyCar race in Boston", "Pixel art off-road race in Fenway Park" and "Pixel art NASCAR race in Fenway park. The Fenway based superspeedways go incredibly hard.
_d5db0d5f-3fea-43b0-b395-a4ae6dfe91d4.jpg
_7786c6d0-287b-4b31-8957-5e04bc644f98.jpg
_8f7022b8-fbd7-4201-9f9b-10198cb5181b.jpg
_9859e668-9978-47f0-8ae1-93dc78ef4c61.jpg
_e62f481a-1d33-46f4-9bd4-8570bf2c4145.jpg
 
The city scape ones are really cool. A sort of fantasy reality / video game rpg map vibe.
 
BeholdBoston said:
ChatGPT / DallE. I asked for an image of the Old State House in 2130. I run a history tour (shameless plug, www.behold.boston) and wanted an image to finalize a series of other Old State House photos I show from the past 3 centuries.
View attachment 46616
Click to expand...

The volume of supertalls in this fantastical image would make some aB'ers incredibly happy, were it ever to happen (it won't).
 
Patrick Winn said:
Amazing. Did you tell the AI to paint it in 19th century realism style? It certainly has that look.
Click to expand...

These ones were just pics I found doing my normal weekly search for Boston. I haven't really played around with anything on my own yet.
 
TheRatmeister said:
Centre St in JP with median running GL service. (And some questionable traffic lights)
View attachment 46619
Click to expand...
The catenary wires and street lights are also really questionable. That's one of the limitations I've found with the way these operate - they can create a general scene but don't have the technical knowledge to make the details make intuitive logical sense. I've used "monster truck" as a prompt multiple times and the suspensions (which are fully visible) are always just random bars going from thing to thing, never anything that would work as an actual suspension. I see the same issue here - it "knows" that there are black lines in the sky over light rail, so it copies them, but because it's trained on images it can't ever know why they're there or how they logically connect. I guess at some point it will get better at such things but I don't know that it will ever get fully photorealistic in that sense.
 
You must log in or register to reply here.

Back
Top