Hand drawn rendering

I like the snow scene very much :star_struck:

after inspecting some details closer it becomes obvious that these pictures are ai generated. definately not obvious at first glance, which is pretty impressive and somewhat scary at the same time.

I suggest the rules for posting pictures in this category/forum should get updated to disallow ai content. obviously this should only be for content that was rendered from geometry made in rhino.


Can you point me to such a detail?

Water not reflecting the actual environment, weird furniture, “failing” perspective/vanishing points and the fact that the interpretation of things like windows and other openings change between iterations - which is also why AI based rendering isn’t quite usable just yet. I’m guessing these were based on just a screen shot of simple volumes (potentially made in Rhino).

There are a few obvious candidates for AI production.

Usually the first giveaway is a “dreamy” appearance of the render, where certain details appear very soft. Sometimes, you get the impression of missing spatial scales.

On the second to last image, there are a series of lights reflecting off the water, only one of which I believe to be “correct”.

On the sixth image, where the sun is behind the house, the nearest water pool has a cling-film like appearance, with reflections that I think cannot exist in this instance. The wall here also looses parallelism with the surface of the water and with its own reflection.

AI also doesn’t really understand caustics, as it is a very computationally difficult problem. As AI is essentially averaging stuff it strips from the Internet anyway, it fails to grasp the subtle detail of water surface reflections. This again comes across as missing spatial scale details, and the water looks more silky/oily.

These images are very good, and I can see the appeal for an architect to use AI, rather than pay someone in-house to “waste” overhead and hardware time on rendering, or even “worse” get charged large amounts of money for external visualisation. I’d imagine selling the concept is probably going to be the aim anyway?

Thanks for the details.
I didn’t notice them.

Interesting to find out how easy I can be tricked :expressionless:

Hi @Charles
I don’t think we’re necessarily “tricked” by AI - at least no more than we are by “traditional” renderings. It is (or will be, in due time) just another way of visualizing, and so far it’s no where near as precise and accurate as a “real” rendering. Based on the specific prompt and the training set of the AI it’ll give you a qualified mish-mash of styles/genres and interpretations. It will, as long as AI’s are “just” super-advanced copy/paste machines, based on nothing but images and their captions, never produce a truly original solution or piece of work - and sometimes that’s enough, hell, sometimes the “look like something I’ve seen before” might even be the better solution. It’s a complicated and fascinating field and it’s moving at breakneck speeds! :slight_smile:

Some details are undefined and more like smeared. Branches of trees are not growing naturally. I don’t know how to translate the German word “Fluchtlinien” to English - but the Fluchtlinien give it away pretty easily. The shadows are not consistent (shadows are on opposing walls). If you look you find lots more that gives it away.
Also the overall composition of the different pictures. A human would typically not “throw around” the geometry or “shuffle” the architecture like this. This is typical for AI presenting the user with different compositions.

Uncanny Valley

Ill second that.


I worked for several years making presentations_perspectives by hand in
the early '80s. (there was no PC yet).

1 Like

Yeah. Perhaps the gallery could have a filter for AI like Artstation does.

I’d also argue It also does no good for aspiring users and modellers to try and achieve and learn if there is nothing here from which to learn.

AI definitely has a place here, and I am curious to see how industry applies it for a “mass market”. But I’d be in no hurry to replace artistry and skill with typing 80 characters. I love learning from others experiences. I’d hate to live in a world where all anyone has to do is sit in a chair, consume, and type into various translations and APIs for GPT and SD all day. What’s the point?