That's the 'built by the lowest bidder' feature. Probably pretty realistic in a lot of places.

Huh, I wonder if they trained it by feeding it architectural renders and "what actually got built" photos...

It's probably just prompt based. Actual fine-tuning for these kind of use cases is getting less common than it used to be.

Likely. You can go into Nano Banana or ChatGPT right now, upload a pretty architectural rendering, and tell it to make it look old, weathered, winter, etc and it will come out looking very similar. Give it an example to really dial it in.

It's not a filter, it's an image editing model

This drink is not a smoothie, it is a blend of fruits and berries.

In my mind "filter" is some specific algorithm that does a single expected transformation

"Filter" is a Tik-tok / snapchat / instagram parlance for any kind of overlay / transformation. It's grown larger than just sepia filters and similar. All the ones that do facial tracking and overlay a mustache or w/e is funny in the moment are also referred to as filters.

See https://www.snapchat.com/lens

There's a pretty clear expected transformation here though? It takes an image and then reduces the "shiny-ness" of it by giving it the same transformation: change the sky to overcast, add material degradation like rust, reduce the landscaping by adding weeds/puddles, and remove the happy looking people.

Also adding random electrical infrastructure and random signs, also removing a statue in the distance in one of the images

Sure, that stuff too. The point still being that it's a pretty predictable set of changes being made to whatever photo you give it.

I'm pretty sure it's either gpt-image-1.5 or Nano Banana Pro in the background, with a prompt like "make it look worn down and slightly decaying".

Right, filtering is the reduction of information while diffusion/generation is creation.

It doesn't have to be a reduction. Swapping the colour channels would be a filter, but it's perfectly reversible.

How is it not just a midjourney prompt? The liberties it takes seem to be better described by 'upload a picture, and AI will be told to make it dingier'. Can't people already do that ad nauseam?

Au contraire, in a rather realistic way

It's GenAI. It does something that's kind of like what you asked it to do, but it will skip some details or add other ones or whatever.

Dreary architectural pictures will be more likely to have electrical boxes, poor materials, etc, so when it moves the buildings from the latent space for cheery bright architectural renderings to dreary wet November architectural renderings, it will be more likely to add some of those details, because that's what's in its latent space.

Don't expect GenAI to be magic.

I have a suspicion that the author of this might have asked the model for those utility boxes.

Yeah - same things I noticed with people enthusiastically using genAI for old photo coloring. Initially it looks awesome, until you realize it can even alter the human face in such a way, that it no longer looks like that person.

My father was really happy with some old photos colored, until I pointed out he does not look like him. Strangely enough he wasnt bothered...

I put in an image and it generated piles of shipping pallets along a walkway.

It also added drainage that would actually improve the building.

It's AI, it makes things up.