Dreaming Futuristic Metropolis

[VIRTUALENVIRONS]

5 hours ago, Tom W. said:

because most times we have a fairly clear idea of what we're looking to achieve

I think with Architecture, this will largely always be a true statement.

 

For those people working in lighting design, especially big shows, etc. this could be a great tool.  Also, Landscape design seems like a good fit.

 

I love this tool, but as an animator I see it in a whole different world, which is where the movie industry is already using it.  I think we are only 5 years away from when Humphrey Bogart appears in his next movie.

 

 

 

[grant_PD]

I don't even see how this works for lighting or landscape design.  Both tend to (as most design fields do) get very specific about details.  AI seems to be at the moment a good "mood board" generator, but I'm not seeing much else just yet.  

[VIRTUALENVIRONS]

As of yet, I have not seen VW's aim this technology at any one discipline.  I think it is so new and exciting that it is hard to figure out where or even if it will land in Architecture, Landscape design, or Lighting.  I don't think it is smart enough yet.

 

But, if I was still in this game, I would be all over this.  It will get smarter, and for those dismissing this technology it may be smart enough to take over their job.

[Jeff Prince]

59 minutes ago, VIRTUALENVIRONS said:

But, if I was still in this game, I would be all over this.  It will get smarter, and for those dismissing this technology it may be smart enough to take over their job.

If your job is generating marshmallow fluff, you should be very worried.  These prompts are basically the instructions I would give an intern or rendering subcontractor to generate some mood boards to identify a direction we might go on a project.

 

But this is typically less than 1% of the effort to execute a project.  Most importantly, it is the most enjoyable part in the creative process.  I don’t know why anyone would want to turn that over to a machine, unless they simply do not have the ability to generate these types of works.

 

Big developers started making everything look the same by replicating original designs instead of making new ones.  Architects started making everything similar to save time.  CAD lulled people into making everything rectilinear and copy/paste architecture.  AI will deliver the cyberpunk distopias we’ve grown to love in movies and novels.  As people stop creating, there will be less for AI to learn from and eventually everything will just be a homogenized blob.  In nature it has been said everything evolves into crabs.  If controlled breeding is stopped, most dog breeds return to what we identify as the common mutt/dingo rather quickly.  It’s not hard to imagine the same phenomena being observed in the creative professions.  Just look at what has happened to common graphic design.

 

All that being said, I’ll stick with doing my illustrations by hand.  Not for the sake of efficiency, but for love and respect for the creative process.  I’de rather have an AI that can convert my BIM model into construction drawings… that’s the place to save significant time/money and a task a soulless program would be well suited.

[VIRTUALENVIRONS]

It is impossible to predict where AI will take us, but I think it will be beyond what any of us can imagine.  

 

Although we cannot see the future.....we can look to the past for answers as the evolution of 3D Virtual Reality was the AI of its time.

 

In 2000 I was charged with reconstructing in 3D the Government Conference Centre in Ottawa back to what it was in 1934, as a Grand Trunk Union Railway Station.  Nearly a quarter of a century ago.  Some of you (students) may not have been born, others just embracing this tecnnology today.

 

My colleagues were all 2D ACAD.  Not only did they not embrace this technology, but actually worked against it saying it would not last, it was only a “flicker in the wind” and I should get back to doing real design in 2D.

 

Well, we know that theory has been disproven and my colleagues although retired, all eventually moved to 3D design

 

This is the Grand Trunk Union Railway Station Ottawa Canada circa 1934.  Built ~70 % Vectorworks and the rest CINEMA 4D.  This was well before the “Send to C4D” command.  It is the original movie from that time.

 

A Long time ago, in a City far, far away...

 

 

[Jesse Cogswell]

I very much agree with @grant_PD.  Lighting design is very detail oriented and the AI doesn't parse Vectorworks geometry.  It doesn't know where hanging positions are or what kind of lighting fixtures you have at your disposal.  Lighting very much operates in the world of reality (as we like to say, physics: not just a good idea, it's the law!), and AI as it stands now doesn't need to adhere to this reality.  The very last thing I want to do is put a rendering in front of a client that looks nice but isn't achievable.  This is already tricky with the difficulty of getting "accurate" lighting in Renderworks or Redshift renders using the Spotlight Lighting Device PIO.  I'm very much dreading the day that a scenic designer presents a rendering of their set made with this tool with all kinds of crazy lighting, and the director turning to me and saying, "I want that.  Make it look like that."  And I know that day is coming.

 

I think that including an AI visualizer like this in a design oriented piece of software is irresponsible at best and down right unethical at worst.  @mjm pointed out an instance where using the AI visualizer resulted in a an artist's signature being present in the rendering.  This introduces all kinds of questions about what this tool is using for training data and whether the originators of that data have been compensated.  I was pretty well horrified when @Luis M Ruiz posted ideas for what was obviously the scenic design of a production of The Lion King done with providing the geometry of an empty stage and generating the rest from text prompts.  In no uncertain terms, the resulting renderings were an amalgamation of other artists' work.  There's already a major argument about intellectual property in the theatrical design world, with a theatre company a couple of years ago getting caught basically ripping off Broadway scenic designs.  Now you've just baked in a conduit to do exactly that, standing behind the ruling that you can't copywrite an AI generated image (but your AI generated image can absolutely be based off of some one else's intellectual property).  Even these lovely sci-fi images that are the topic of this thread contain prompts strictly calling for specific artists' styles.

 

This all being said, I actually do want to see AI integration in Vectorworks, but I'd rather it automate the mundane and repetitive tasks rather than the most enjoyable and creative part of the process, as @Jeff Prince alluded to.  I would love if I could tell VW to dimension a Viewport.  Or if I could tell VW to make sure that all of the symbols in a folder have geometry set By Class and in the correct class so I don't have to go through them one by one.  Or have VW add a text label linked to a record at the center of all selected symbols.  Or add keyed notes to selected sheets and make sure that the legends all match.  Or go through all of the selected objects and vary the horizontal and vertical offsets of this wood texture so that they don't all exactly match (or better yet, build some decent non-repeating wood textures).  All of the crap that we spend hours needing to do by hand instead of getting to actually design things.

 

I do also see some use with the AI visualization engine, but mainly when you provide a decent amount of geometry and drop the Creativity slider a bit.  @Luis M Ruiz produced some stunning images by providing a hidden line drawing and asking the AI to make it look like a charcoal sketch or an oil painting.  I think using the AI as an extension of something like the old Artistic Renderworks options is a fantastic idea.  I talked to an exhibit design firm about this at length a couple of weeks ago.  They were using an external AI visualization engine that they had trained using artwork from designers from the firm.  They had found that a lot of clients were turned off by 3D renderings early in the process, since they believed that once it had been committed to a 3D model that it would be too difficult to change.  So they started to make fairly simple hidden line geometry then fed it through the AI to make it look like their own "sketchy" style.  But in this instance, they had strict control of what the AI was using as training data, something we don't currently have in Vectorworks.

 

It really does feel like Vectorworks is chasing a trend here, but believe enough in it to risk their software for it (VW2024 Update 4 anyone?).  I'd much rather see stacked wall components, a stair and railing tool that doesn't make me want to pull my hair out, finished cabling and power planning tools, and a copy of Vectorworks bug free enough that I feel like I can safely adopt it the year it releases.

 

PS - Apologies to @Luis M Ruiz for standing on a soap box in the middle of his thread.  The images are indeed quite lovely, but I do question the ethics behind them in a design oriented piece of software.

[mjm]

@Jesse Cogswell—"…the day that a scenic designer presents a rendering of their set made with this tool with all kinds of crazy lighting, and the director turning to me and saying, "I want that.  Make it look like that."  And I know that day is coming.…" It has already come for me.

And it was as unpleasant as I expected.

[mjm]

OMG.

"…All of the crap that we spend hours needing to do by hand instead of getting to actually design things.…"

[Matt Overton]

On 5/13/2024 at 8:46 AM, Luis M Ruiz said:

Prompt: Oil painting of futuristic city with tall buildings and spaceships, blue sky, gray tones 

Is this a city of constructed iceberg buildings?

[Luis M Ruiz]

I have this file that I've been growing since 2016. Lots of friends provided massing models. We use project sharing. Recently a new AI plugin was introduced and I am curious to learn how it can be incorporated during the design process. I find this particular plugin requires geometry as a starting point. A shaded white model with minimum colors works wonders in this case.

Prompt: Spiraling buildings and skyscrapers with a circular design stands in a city skyline. The buildings are covered in greenery, glass and steel, cinematic lighting, water front, dark reflections.

Geometry attention at 0 > 50 >100%

[Luis M Ruiz]

Why local when we can do Mars?

Prompt: Futurism style, city designed on Mars in year 2500, red dusty sky
Negative prompt: Moons

[VIRTUALENVIRONS]

Hi Luis,

I was thinking, what would a city look like in the clouds?

[Luis M Ruiz]

I decided to shift to a more artistic style. Maybe for a t-shirt

Prompt: Side view abstract artistic rendering of a futuristic city skyline. It consists of intertwined, curved white lines on a black studio background. The soft and flowing texture, with the lines creating an ethereal effect, golden dust.

[VIRTUALENVIRONS]

Such a beautiful image.  You could sell that for a wall mural.

 

 

 

 

[VIRTUALENVIRONS]

"Dreaming Past Metropolis"

 

 

Street View

 

 

[Luis M Ruiz]

Are these AI generated?

[VIRTUALENVIRONS]

35 minutes ago, Luis M Ruiz said:

Are these AI generated?

Hi Luis,

The Suburb is all Vectorworks geometry.   I used C4D for the city, just used there stock models with luminance textures for the windows.  C4D for the trees and lampposts.  Rendered/animated in C4D.    I created the original model in 2012 for a video on waste disposal (see below).  I  re-purposed it in 2015 for a night demo.  The 2015 image is the one I posted.

 

It was Sunday, things are slow on Sunday.  I thought this might make for some discussion on AI vs VR or how to integrate the two.  If AI could do renderings like yours, but on VW CAD geometry, it would be a game changer.

 

 

[BartHays]

I think @VIRTUALENVIRONS is looking for is an AI system that creates images based on actual 3D geometry. Whereas today's AI visualizer is simply a 2D image to image generator that can use a Vectorworks view as the initial image

 

I think there might be an interim step. What if we could use AI to automate the Massing Model tool?

Describe the Massing style you want "Suburban neighborhood," or "a futuristic city in the sky", then define a bounding box. The AI would generate unique structures (3D models) in the style described to fill out the site. Maybe one day we can have a reliable Prompt to 3D model generator that can take on tedious modeling tasks. 

 

 

[VIRTUALENVIRONS]

@BartHays  Yes, essentially unless VW's can somehow link a complex model to AI Rendering, we are looking at two different programs within one application that don't have any connection.  it is not like Vectorworks and C4D.    I know there has been some talk about how this will help in the early stages, but it could be more disastrous than helpful.  

 

 

 

 

[Luis M Ruiz]

Based on the image provided, I was able to create a few AI images.
Prompt: A sprawling city skyline adorned with majestic towers and soaring skyscrapers, juxtaposed against a quaint row of charming small houses. Illuminate the streets with a captivating array of lights that dance beneath the starry night sky, a realistic and cinematic scene

[line-weight]

I think in the foreground it has created a kind pf prison camp where architectural draughtspersons, renderers and the like will be incarcerated once the AI overlords deem them entirely redundant.

[Todd McCurdy]

Maybe a whole lot of mature trees would help with the Stalag vibe

[VIRTUALENVIRONS]

Hi Luis, so cool, buy perhaps @line-weight has a point.😀

 

So, some progress here.  If I was proposing this design 13 years ago, I would use your image, knowing the CAD model was in place.  

 

Kudo's to you.

 

 

 

 

 

[VIRTUALENVIRONS]

@Luis M Ruiz Hi Luis, dialling in to what the Visualizer needs, below are two images.  This is the legendary AVRO Arrow.  If you decide to see what you can do, make sure to use the prompt "Avro Arrow".  There won't be a problem getting info.  Or just look up AVRO Arrow.

 

That is a large four point crane in the background.

 

So, I made two exact images, one the way I wanted it to look, but thinking you can make it look better.  The other is just a black and white for you to work from.

 

FYI, except characters, all modelled in Vectorworks.

 

regards...Paul

 

[Jeff Prince]

@VIRTUALENVIRONS I'm failing to see what your posts have to do with the topic at hand.

Maybe start a new thread about rendering your aircraft, could lead to some interesting techniques.

Don't you have access to the AI Visualizer to try something on your own with these images?