It's interesting to me, the model does not know what a house is like, it knows what pictures of houses are like. So it does a good job of making pictures that look like pictures of houses. But if you look closely, a lot of the details are really weird, unbuildable, or just non-sensical.

All of the image-gen models have this problem - look at the hands and faces in the generated images of people and there are often bizarre deformations.

It's fascinating because it's the opposite of how children learn to draw. They tend to think about the pieces that make a thing and then try to put all the pieces on paper, and they end up making a drawing that (for instance) looks nothing like a person but has two eyes, a nose, a mouth, etc. in roughly the right relation to each other. (They rarely draw ears though!) The child is thinking about "what makes a face a face" and then trying to represent it. The ML model is sort of distributing pixels in a probabilistic way that comes up with something very similar to the pixels in a sample image in its training set, superficially much better than a kids drawing and yet in some ways much worse upon close inspection.

perspective and shadows in this one are particularly avant garde

https://thishousedoesnotexist.org/4036194

> a lot of the details are really weird, unbuildable, or just non-sensical.

You would see a lot more of these on the images that didn't make the gallery.

I don't know if it's just me, but it seems to only generate luxury homes in a contemporary minimalist / modernist architectural style. Not once did it generate anything resembling a Tudor mansion or Victorian cottage or Midwestern rancher or split level.

Nothing really wrong with that and it's neat anyways, but seem to represent its own sort of strange bias in a sense.

If it was done text-to-image, what you see is the result of using the same prompt.

E.g. in Stable Diffusion, from a single prompt like "modern residential interior, concrete, glass, wood", you can generate a vast number of images, using different seed values.

The images will have something in common due to the prompt.

The Tudor mansions are almost certainly there in the training data, but not being selected for.

It does seem to love lots of glass and flat roofs, doesn't it?

Neat, but feels like it was trained or prompted for luxury homes. I clicked threw a couple dozen images but everything had that style of multi-million dollar mansion/condo with a pool and only glass for walls. Nothing even remotely suburban or rural. Curious if this is due to the prompts (assuming this is SD sourced) or due to the training of the AI model.

This is the intention: https://twitter.com/levelsio/status/1566887688241815552?t=8y...

Also:

https://thisxdoesnotexist.com/

https://this-person-does-not-exist.com/

I feel like we are at the tipping point where the ability to generate scenes/faces/houses/art that can reliably fool human senses becomes commonplace.

What? None of those look like real houses... It looks like video game quality at best and almost all of them have some aspects that make no sense at all.

I'm waiting for "this AI does not exist"

Welp, here you go. https://thisaidoesnotexist.com/ It was on HN a few months back.

Noob questions: Are these images that cheap to generate (once the model is trained) that we can get a new one fast on each browser refresh?

Are all the _this x doesn't exist_ generators based on similar scripts, only the training datasets differ, or are they somewhat tuned for the specific domain?

Generate time is a touch too high right now to do it live unless you're userbase is exceptionally patient but you could run a batch process across a bunch of machines for a while to achieve a similar effect.

I'm waiting for the crossover with McMansion Hell where somebody uses similar image generation to generate the exteriors and interiors of McMansions.

Judging by recent generations of McMansions in my part of the U.S., that technology was "perfected" some time ago.

I'm waiting for Zoning Boards, Planning Commissions, etc. to get their hands on that tech. - and (BIG ASK!) outright ban construction of anything which looks like it might have been generated by it.

Not sure if my experience generalises well, but a few of those houses generated were impossible to build.

Indeed, https://thishousedoesnotexist.org/assets/houses/3027276.png reminds me of https://en.wikipedia.org/wiki/Impossible_cube

What am I missing? It looks a bit whacky and expensive/James Bond villain lairy, but what's inherently unbuildable?

Look carefully at the perspective.

Yeah same, more a collage generator than an architect.

It's less convincing than the other projects like this, wonder if it's a uniquely hard problem or a weakness of this particular implementation.

That's not all that uncommon with human architects, either, at least if you go by what the contractors say.

I refreshed ~10 times - all images had escher-like impossible elements.

Just a gallery of AI-generated images on a theme. Brace yourselves for a deluge of these.

Seems to have been trained on CG renders of houses instead of actual photos of houses.

the next level would be the A.I making "sense" of the interior spacing or the layout in general so it could generate more accurate houses

Even better than making sense of it, but just give the AI all of the parameters that you want to build or remodel a house and have it give you a 90% complete model that an architect/designer can finish off, reducing overall cost and time.

In some parallel universe it does