Tudor Vasiliu: AI For Architects, from Prompt to Art

We talk to Tudor Vasiliu, founder and director of Panoptikon, about how architects use AI to elevate visualization without losing control, unpacking “AI passes,” professional guardrails, and why speed still needs expertise. A live demo shows rapid mood and lighting iteration, while we call for better client tools and more usable AEC software.

• AI passes that enhance materials, lighting, foliage, and people
• Why expert workflows beat one‑prompt myths
• Accuracy and ethics in public review visuals
• Faster iteration with in‑house tools and cloud models
• Video enhancements from stills and CG using diffusion
• Client expectations for immersive, truthful previews
• Bridging 2D habits to BIM and lightweight planning
• Startups’ agility vs software giants’ slow pace
• Natural interfaces as the next AEC breakthrough
• Guardrails to keep framing, dimensions, and intent intact

Transcript

Roopinder 0:39

Hi nice nice to meet you as well. Where are you calling from?

Tudor 0:44

Uh basically Bucharest in yourRomania.

Roopinder 0:46

Oh, you're you're quite far. What time is it there for you now? It must be the evening.

Tudor 0:51

It's it's uh 7 p.m.

Speaker 1 0:53

7 p.m. Oh, I'm keeping you from dinner and family probably.

Tudor 0:58

Almost there, but uh I I do have a late schedule. Founders need to leave last, eat last.

Roopinder 1:07

Oh yeah.

Tudor 1:09

Right. It is like that, but I I do enjoy it. I mean uh I tend to sometimes work late and get up late. So it's a bit of uh I have a different schedule here. I I have several projects uh in and out of the of the studio, but uh it's an architect thing, I guess. Uh workaholics.

Roopinder 1:33

Workaholics, yeah. I d you know, I I don't detect a total Romanian accent, not that I know what one sounds like, but uh where were you educated? Were you you have uh English? Oh thank you. Yeah.

Tudor 1:47

No, I um I did uh get into uh studying languages very young, so my my parents were very keen to have me, you know, study French and English, went to a French school, learned French uh very young, and I think English was was actually first. So I had good tutors, I had uh I had uh university teachers and people who were you know super passionate about languages and especially English, and I read uh read through you know English literature and uh it's a passion. Yeah, yeah. So I see that uh I I do I do have an incident with all our clients across the world and the US, so uh it is it's I think I was lucky.

Roopinder 2:37

Yeah, yeah. They say you pick up the accent of your tutor uh out of time.

Tudor 2:47

She had this American accent, and uh I did start with uh British English, London accent, pure brief, and then I just switch and I never looked back. I I just love it. Yeah, yeah. But English is it's it's a it's a huge amount, it's a huge vocabulary, it's just fascinating to see you know uh novel and written English and ancient English and modern one that is uh intense in a way. I I do I do appreciate how how English speakers just uh they so powerful thoughts with just a few so few words.

Language, Education, And Global Clients

Roopinder 3:24

You you mentioned your international clients. I noticed uh one of them, your work by the way, uh in that uh Naples facility was amazing, I've got to say. Yeah, that's uh stunning. That was so you probably have you probably on the on the schedule of the US quite a bit at the time.

Tudor 3:44

Um yeah, and they I mean I I prefer their mornings for calls, uh so I so I do get some sleep. And they uh they're also uh very understanding these uh these people and very kind. So uh with a gentle team uh for you know for every client understanding it's a breeze. And uh fact is we deliver uh by our evening, so by their morning, it's in their mailbox, which is great for them to just stick up the views and the draft views and the interim proposals that we're making. So Fables was uh yeah, it's a it's a massive project and one of uh projects that got us uh uh a lot of attention.

Roopinder 4:28

Yes. I mean that they're just stunning images. I have to look very closely at them, and I'm a little skeptical. I always think, is this a photo? Is this how this is this faked? Because down to the details, the the leaves on the trees and the it's it's just amazing what you what you've been able to do with that.

Tudor 4:47

Those are all that's all rendering, plus AI passes um at the end above the the final image. So we're using all sorts of tools nowadays to improve the results and uh what you're seeing there, and especially the the bark on the trees and the leaves and whatnot. That's an AI pass that uh adds detail and um sort of makes everything look twice as spectacular or four times as well.

Roopinder 5:18

Oh, I see.

Tudor 5:19

So although the quality the quality that we're trying to put in an in 3D, I mean the quality standard is there, but you cannot rival AI nowadays, it's just too efficient.

Roopinder 5:32

Yeah. I'm glad you brought that up, AI, because no conversation these days is complete without talking about AI. But how did you what is what exactly is an AI pass? When you say you take a rendering, is uh is an AI pass? How does that how is that used and how is that how what effect does that have?

Inside The Naples Project Visuals

Tudor 5:52

Oh, definitely. Um so essentially what we're focused on uh uh with you know producing visuals is just uh engines like uh Google's latest Google Gemini 3 Pro, but all the other open source models that uh architects use to generate images or to improve images, um so we have been trying to use as much as possible every possible platform out there. Uh we do have a ton of subscriptions and testing everything. So the holy grail is to get these results into production and to have your product look as good as possible at very high resolutions. And yes, the the AI pass, how I call it, is you take your image, you uh pluck it into a um one of these platforms or local um, let's call it stable diffusion, was one of the original um AI diffusion platforms that use um AI models, diffusion models, image diffusion models, as they call it. Uh you pluck it in there and you use a sort of workflow that is trained on similar images, and then at the end of this workflow, you get, um by workflow I mean the set of instructions that use some pre-trained uh AI models. And by the end of your workflow, uh you can be uh getting you can get uh an image that looks um way, way better because you're trying to improve, let's say, the definition materials or the lighting or the people themselves. People are a very um hot topic in uh in the AI generation for images at least. Uh, you know the problems with uh getting more than five fingers or getting some deformed faces. So these models, they you can plug in there uh different uh uh checkpoints or different AI models that take care of bits of these images. So they sort of deconstruct it and they can understand the depth in the image, and you can work on on these layers of like um integrate fog or create depth in the shot, like the uh romantic paintings. You know, there's a haze in those images. Or you can just um selectively improve characters to provide them with more detail or redress them or improve the faces or the features. There's a lot of there's a lot of options nowadays, and um I can talk forever about that, but um I um it's something that most studios nowadays are using, so the standards is the the bar is so much higher now than it was uh than it was two years ago, let's say or four years ago. Um and this regards also images but also video. So video nowadays can benefit from this um treatment with AI, which was not there before, and um you can even generate uh images, uh sorry, video based on images, like first image and last image, you can have them rendered in 3D, but then integrate, put that in your in your preferred AI tool, and uh it will generate the motion, but you can also integrate characters inside um these AI generated videos. You can also improve your own CG 3D generated videos with AI. And you can run these models, AI models, locally if you have a very strong, very powerful GPU or set of GPUs, or you can uh run that in the cloud. There's so many uh platforms out there that allow you to run uh either open source open source models, uh AI models, or proprietary models that other uh creators have uh have produced. Uh some people train their own stuff. Um you know it's it's a I think most people get stuck at you know the amount of boxes that they have instead of uh you know if you get recommended by someone, use that, you're probably gonna end up testing that that very tool and seeing it seeing if it's for yourself or not.

Roopinder 10:13

Yeah.

Tudor 10:14

But uh what we do, generating image imagery, you know, high-end uh imagery that needs this sort of uh strong art direction, creating, you know, conveying these ideas and these stories into something that really tells the story of a project, uh, that has been really time consuming. Now it's so much faster. It's still the same kind of work. You kind of need to put in your sensibility and knowledge as an as an art architect and as an artist, but it's so much faster nowadays with the help of AI.

Roopinder 10:52

I believe that it's faster, but on the other hand, you've developed a very sounds like a very sophisticated process to generate the final image, which is very high, again, very high quality, very realistic. And that that's interesting to, I think it'll be interesting for people to hear that it is a process, that this is not just a simple prompt that you can just say, hey, ChatGPT, for example, generate an image from this CAD drawing. It's not like that at all. You've you've taken the you take the picture apart, you work with it in pieces almost, you work with several aspects of the picture, you maybe you work with a timeline. It's quite a process, actually, and that you've developed. It's not a simple prompt, definitely not just a simple prompt that anybody can do. It's not far from it.

What An AI Pass Really Does

Tudor 11:44

And um I appreciate you bringing that up because um we've also we've already had had this sort of conversation back in the day when teachers in in university um uh were referring to whatever work was done by computer as a uh a cheat, as a cheat trick. You know, you push a button, you render the image out, and you have it. Where's the you know, where's the hand drawing, where's the uh the um you know the painting or the watercolor or the brush? Where do we see that? Well, where do we see the the the the tens of hours put in a in a masterpiece, beautiful rendering of a project, and we were like uh it's not the case. I mean, if you need to draw your project in in AutoCAD and later in BIM software, you're not just gonna get it by by uh saying a few words or by just clicking a button. So I think you know, seeing the acceleration of our work is performed since the 90s when I uh when I went to university in 97 and up to where we are now, I think it's you know it's just an exponential curve, but it's the same idea. Behind it, you need to be an educated specialist. Uh I am an architect, and uh, you know, I I employ architects only just because um these projects are not just kids playing with, you know, you cannot be someone that's irresponsibly generating images when you don't understand the architecture and the design and the all the people involved in creating this. You have engineer, you have the developer, you have the architects, interior designers, landscape designers, traffic. You know, there's so much that comes into these projects that we have to be aware of. And the way we show these um these projects, whether they're in context or not, it needs to stick to the reality of things. And I think, you know, I I was a bit lucky because my at the at the beginning of my career in in the 2000s, so I I essentially I've always done uh commercial work for various architects, even family. My my dad was an architect, and I did some renderings for him or my uncle back in the in the actually in 2000 and whatnot, but what I was was act what I was actually um lucky with is to start my career working for developers, real estate developers, and not just uh people you know doing napkin drawings or idealistic architects, but people who needed these um these uh uh specs to be as accurate as possible. And it is a process where you have to meet the project, you've got to meet the client, you've got to listen to them and understand what their pain points are and what they're trying to achieve. And um take it from take the drawings or take the 3D model and go through numerous revision rounds. You've got to propose, we what we do here is explore the project as much as possible. We're like virtual uh photographers or videographers most times. We are free to change whatever we want. We can change the time of day, the seasons, the position of the camera, we can go into space. So it's it's really um it's a freedom that a photographer doesn't have, and we don't have to spend weeks on sites to get that multitude of views and to show the life cycle of the project, in fact, um in a set of renderings. Um it does uh come with uh, I think, responsibility, the same responsibility that a designer has sometimes, and even more because these visuals end up published and analyzed by city hall, by neighbors, uh, for example. In some states and in some uh countries, neighbors can faithfully oppose a design if in any way you know shows it that it neglects some aspects of their life or of uh regulations in that area. So we've had cases that showed us that we have to be extremely diligent when it comes to what kind what this work shows and how it shows it. And I think blaming renderers for showing idealized versions of these projects, and uh um it's a it's a problem, I think.

Fixing People, Depth, And Atmosphere

Roopinder 16:38

It's a big problem, I have to say. Um I think it's gonna be an even bigger problem as we move on because I'll tell you a story. It uh involves an office neighbor of mine who I hope can't hear me right now. But what he decided he was going to ask ChatGPT what a remodel of his house looks like, and sure enough it gave him an image, and it was like, wow. Right. Like that, oh, we can do that. And but he kept pressing it for additional changes, and before you knew it, he uh it was starting to get really weird. It would be windows would move, doors would move, roofs would change. He was drifting from the original concept because you can't really control it. It this is AI applied very by an amateur. But uh, and I think you have to fight that concept because AI applied by a professional is much different than AI applied by an amateur. And but people are getting the impression that anybody can make beautiful renderings, right? And they're not grounded in reality. You had the luxury or the privilege of working with a developer, so you're very grounded with what residences and offices and shopping centers need to look like. You're not going to make fanciful cave dwellings or towers in the sky or something like that. Those are crazy renderings. And this isn't just architecture, but engineers have to do have have this problem too. Writers have this problem too, right? Everybody everybody thinks, oh, make me a house. AI, make me a car. AI, make me a right, AI, write me a story. That's right.

Tudor 18:22

That's right. I mean, in history, we've had architects who were bold. Let's say, uh, I don't know if you're familiar with uh Lebbius Woods, he was an American architect promoted deconstructivism. Uh-huh. Um, just incredible exploded structures made out of um pieces, just I mean, this guy was experimenting like crazy in the in the 70s, and some of his work had was built, but he was an architect who was not just anybody off the street. Um you have Zahadi, for instance, and um Frank Gehry, uh, also incredibly creative architects who challenged um actually every norm in what you know organic architecture can mean and can can uh can be constructed like. And studios like uh Gehry's studio have been developing um procedural tools and probably AI tools for many years now to tackle with their with their challenging designs and to make the the work of um structure engineers and all the others easier and to be able to take that into construction uh without losing control, without relying on a third-party engineering firm, for example. So it it's also this also has to do with you know the boldness of what designing means. And I think once you see these AI generated um examples, they're drawn from you know train, they've been trained on existing imagery. And of course they kind of pushed boundaries, but it's still, you know, if you scour the internet, you will find non non-AI generated images at some point showing like a building in the form of a shoe or of a fish or some organic type of thing, you know. It were originally maybe it was a drawing, someone drew something that was impossible or very expensive to to build, right? Right. Unless it's loading and it's still not there. You gotta do something like uh anti-gravity systems. But um people have always tried to push these things into um into something that's that's uh way beyond what's possible. And I think you know it's you just gotta be careful who you're hiring to be your architects. Um because once they push the envelope, the push boundaries, uh it's it's gonna cost you. Yeah. You can probably uh you know, with a with um you know a bit of uh careful uh approach, you're gonna be able to build that, but it's gonna cost you a lot. So um, you know, you do see these uh public buildings everywhere. Um you know uh their threshold is billions. You know, some buildings cost uh buildings, up to two or three billion, some of them. Um not just public buildings, but also, you know, uh I mean public buildings mostly. Hospitals are expensive, airports are super super expensive because of the system inside there. So um I think it's always about the diligence or the responsibility of the person uh employing that tool, and us as at the end of the day, designers, we do need to strive for control and have. In uh some guardrails for this um, whatever tool we're using, whether it's ChatGPT or a or a professional tool that someone develops for other professionals, um they need to have realistic expectations and uh guardrails set um deeply within the generation process, so it's not uh so it's a helpful you know result, not something that uh hinders the efforts of other specialists. So you've taken care and you're hiring to hire only architects. I imagine other architectural firms are probably not being as careful and the uh then they're hiring uh I don't know, artists or somebody off the street that knows how to use AI, perhaps, but I I suspect they're gonna have to spend just all the money that they save and use that for checking the checking the uh drawing or correcting the drawing or correcting the uh image or the concept. I I don't think there's gonna be any savings there, and I think it's gonna be a backward step for them. Are you encountering any of that in your profession that uh architect firms are using uh AI too much

Speed Gains Without Losing Craft

Roopinder 23:18

or too badly?

Professional Process Vs. One-Click Myths

Tudor 23:20

That's a great point, Lucinda. I think um you know, my my profession, I'm kind of a fringe architect and just working with visuals. Um I do encounter it in in our industry um uh a lot of uh people who haven't studied architecture. And I do trust that you know the passion can make and drive you to learn uh a lot just not going to school. Architecture is so difficult to learn. I was in university for seven years. Um, you know, some some uh some do their uh PhD studies further down the road and spend a lot more time in in school. But it's a lot it's a wealth of information to go through, to crunch through, and um who whoever is is not an architect has to has a lot to cope with. Um so that is one. The other part, whoever's using AI, I think you know, the the judge of the result is the um the client, the end client. And typically, if that's an architect or um or a developer or um you know someone who's um astute in understanding what what they see, uh they will be the judge of uh what's happening in that image, that animation, or that um, let's say, uh virtual uh experience that they're um experiencing. Firms like ourselves are very careful to manage this process, the creative process, and self-checks, self-correct before the client ends up seeing anything. So in our creative industry and in architecture as well, I think you know, revisions are are the ultimate control process that you must have in place. And if you're hiring someone who's not an architect, then I I've I've I've seen, you know, one of the one of the best studios out there that I that I could even mention without bringing the sweat is called Mir. MIR. Um it's a Norwegian studio. Uh these two guys, the founders, are not architects, they've been uh they they studied uh arts, but they are so uh uh good understanding architecture and probably hiring people that they um you know their images are so um infused with with an artistic sense and what what you get, I mean, um it's just something that is it's always unexpected, and you know, like the one of like one of the if you think of one of the high-end photographers nowadays in the world and in whatever fashion or nature, they're one of the high-end photographers of of virtual architecture. And I I'm honestly I'm trying to remain unbiased. People can learn, uh, you can either go to school or not go to school. You can have that drive to learn new things, and um I I'm pretty sure you can you can do it. But if you want to get hired as an architect by a by an architecture uh design firm, you're probably not gonna make it unless you someone gave you a shot, and uh, without a degree, you ended up either designing yourself, um getting some experience under your belt, and proving, getting those proof points in your CV that you can handle complex information, you can understand uh, you know, all kinds of uh design processes and uh specifications and whatnot, and put them at work into uh design. You know, our our industry, our job here producing visuals is far softer than any design process, but it still involves us, you know, we're doing a lot of interior design, for example, for projects that don't have an interior design or they bring us these empty spaces and say, we have to sell these apartments, or we've got to sell this building and produce some visuals for this office building or for this public building because no one is there to focus on the details. And our designers would jump in and uh follow those steps and you know producing mood boards and showing them to the client and showing options and then conveying that into a beautiful image. Um they have to know, they have to be aware of what's uh what design means of uh all the trends and uh functionalities. You gotta uh there's famous books. Uh there's a book book by Neufer, who's a guy who put all the standards of living and working into one single book, and that's the Bible for whoever wants to create a functional architecture. It's your Bible, it has been adjourned, I believe, and all of it. You can become uh a responsible designer, uh, even if you didn't go to architecture school.

Roopinder 28:39

Your visualization game was probably quite advanced already before AI came on the scene. You were probably doing uh ray traced images and and uh graphics, right? And then you jumped on AI probably sounds like as soon as it was available, you saw it as a as a tool to enhance or or uh maybe even um accelerate your process. Um so that that's really good. Do you find that the that the need for visualization, advanced visualization, I'll call it advanced visualization, 3D rendered images, has now taken a big jump in clients' expectations? Like more clients need to see that now than ever?

Responsibility, Accuracy, And Public Review

Tudor 29:25

I you know what whatever you see on the website has zero AI in it. So that's that's a standard that I mean um Naples was a project where we did get uh that perk. We did get the chance to improve those images before we we handed them in. But that's one of probably one of the first because you gotta put in the time, you gotta wait for the machine to render that that stuff, or you gotta pay for a service, a third-party service. And I think it's inevitable for whoever's looking at these uh products to expect the best. Once you you're used to potentially outsourcing your work to a studio like ourselves, um, which I believe is gonna get uh less and less outsourced, that sort of work, because AEC tools were probably gonna get their uh own uh AI generation generated uh models to you know create an image from from uh from a sketch and whatnot, getting they're gonna get more and more precise. But the standard comes from inside the industry, from the design industry, from the visualization industry. Everyone's pushing for the use of AI to accelerate and improve. So it's not just you know cutting time. Um I honestly believe that we're not there yet. I mean, we cannot deliver final work just you know, with just by pressing a button. As you said, all the work behind there is just manual labor, conversations, meetings, uh going to the site, photographing materials, and implementing everything on stack. Um, but everything that we see nowadays is gonna get some scarabs pumped into whatever the results are gonna look like. And you're gonna start seeing, I mean, you already start seeing Hollywood great um videos generated by use of hybrid workflows for bespoke projects. I mean, it's not just about prompting and giving your your AI tool just a set of directions, but also starting with your project, with your design, customizing that, building something like a workflow that you can possibly then reuse that will generate amazing results. We have produced our own, let's call it, rendering AI rendering platform in-house, and we've been using like crazy for the past few months, and um use this Google Gemini, but I I know that a lot of design uh firms out there have already, you know, put out their or they've built their own um power as you know, software powered by AI or by machine learning. There's a huge uh firm called Foster and Partners, probably the number one in the world that um has also been uh mentioning their internal tools for a while now in the lectures and showing what they can do. So I think it's just you've got to look at these at these tools and understanding how they can help you and how they can help eventually your clients and um employing them sooner than later because um everyone's speeding up their work and delivering more and more interesting and more diligent, and um nowadays not using AI is probably ten steps back than all the other uh people out there. And this tool that we build is helping us iterate from the initial steps of any project, we can get creative, whatever you can dream of, you can just um use a sketch of your of your project or of your client's project and you know start thinking in in our direction, our director uh terms, and having your ideas in in the form of a beautiful uh crazy, amazing image in seconds, and you can potentially generate a myriad of directions of you know, visual directions for just one shot or several shots, and then have your client uh decide which one would would tell the story of the project best. And that is something that you you should have probably spent you know days, weeks um trying and failing and you know um decanting the the best one, the best shot. Uh not anymore. And you can probably not deliver that as a super high resolution rendering for print yet. And you still need to be to be using your AI software to generate that high-res image and you know, meet somewhere in the middle, but it is a giant leap compared to uh the way we did work before.

Roopinder 34:44

Yeah. I'm reminded of uh I was at a NVIDIA conference and they showed, this was a couple of years ago, where they showed how uh NVIDIA had software that would let you finish the last bits, the last, let's say, let's say 10% of the image. Rather than let the rays bounce around endlessly and do the final image, AI would jump in there and say, oh, this is what this is an object that it recognizes that's finished this off, or these pixels are enough like the other pixels that we can finish the job here. We know where this picture is heading, and we'll just add the final details and make it crisp and snap and and uh it sounds it seemed pretty effective to me. I don't know if you're using Nvidia's software to do what you're doing. And then the other thing I wanted to ask about another software, which I'd like you to comment on, is SketchUp. I'm sure you know SketchUp quite well. Absolutely okay. Now they introduced a couple a year, I think a year and a half ago, they introduced a, it was their first implementation of AI. And basically you could give it a sketch, not a hand sketch. Let's say you could give it, okay, you could give it the black and white gray image of a of a building, and it would render it. It would give you materials, it would give you shading, lighting, all that stuff from simple prompts. Now that leads me to believe that more people, I think you sort of intimated this that more people are going to start doing advanced type rendering, advanced looking rendering on their own, right?

Tudor 36:23

Yes.

Roopinder 36:24

Okay, right. So that that how do you counter that? Because does that leave you with, okay, let them go ahead, let them have a shot at it, but when they really need professionals, uh or it's a really big project, we'll take over. Is that where you see your firm fitting in?

Tudor 36:42

Like that that is a fantastic question and uh question, and when you look at your demise, you look into the future and you don't really understand what's going on. I think everyone's scared of um of you know what's past AGI. What's past AI, not even AGI. I mean, uh AI is becoming so advanced that we don't need to reach AGI for for every everyone out there to start using AI to do something and possibly replace people. Um I think, you know, I I I know how design is is getting um uh uh produced, and I do think that there's so many steps that humans need to uh need to undertake to create the designs that we're uh seeing getting built. I'm not speaking from my own perspective because visualization is just something that it it may be, you know, it may be just a commodity, maybe embedded in in any AEC software in the future in five or ten years. Um I'm seeing, you know, just by looking at at your podcast and seeing all these amazing uh founders and uh people doing, you know, working on these AI-powered tools, there's so much in there, I mean, way uh more advanced visualization. I mean, just so cleverly, so clever use of AI inside all sorts of uh processes. But I I do think that there's gonna be, I mean, AI helps um at uh will be helping at any stage, I I believe, uh, not just visualization. Um and from our perspective, I I think we're uh we're I I I'm I'm always seeing people who are not even BIM ready, designers. Yeah, quite a bit. Because they love doing their hand drawings and maybe getting into AutoCAD, old school to the AutoCAD, I know. Have we had the resources to invest in, you know, training or it's still Steve Kerbso learned to use MicroStation, uh Revit, or Arcika, or whatnot. So back to works. Sketchup even. SketchUp is is not uh not an easy uh software to learn.

Roopinder 39:15

No, it's not.

Bold Design, Costs, And Guardrails

Tudor 39:20

So speaking of that, I think what do you think of BIM 2.0? Because that is something that, you know, I think it's a bridge between completely illiterate to being very BIM literate. Uh-huh. But there's this gap. And I think all these people creating, like our call, I've seen uh the presentation on your on your um uh on your blog, if I'm not mistaken. Yeah. Uh or there's so many other tools, and I have a I have a list of all of them because I'm actively studying them. Uh-huh. I want to help, I want to help these illiterate architects perhaps recommends a tool that can at least have them jump, have them bridge this gap and going closer to BIM because it's going to help them a lot. And I'm seeing these tools, like, they're so I mean it it makes so much sense for them to exist. Yeah. Doesn't it? Yeah. I mean, Fierce, what do you what do you think of of BIM 2.0, like hard calling tools like that?

Roopinder 40:24

I you know, they exist. I think I think they're interesting, but I think I think I think exactly like what you said, I think there's there's a still a lot of technology architects in general have not implemented. And for them, I think they're gonna they're able, they're going to try to make a bridge too long to get from they are now to what they see in AI. And I think they're gonna not understand the fundamentals of 3D parametric design, BIM. In other words, I think they're that they're gonna try to jump over that step. There's a huge p population of architects that are still working in 2D AutoCAD, especially in the States. It's just like, wow, have you guys not got out of your studio effort? And have you not seen the world? At personal experience, we had a we were remodeling our house. By the way, I'm in north of San Francisco. I'm just practically down the street from Autodesk. I write a lot about what Autodesk does, all their AI implementation, I see all their great visuals, and it's great stuff. And uh and and I and have the architect come to the come to our house and he gives us our he shows us the plan of what he's going to do with our house, and you know what it is? It's 2D, it's a blueprint. He enrolls a drawing on the kitchen table. And okay, I am I'm trying to 3D image that in my head. I'm fairly good at that. I've done that for a long time. But my wife is looking at that, it's like Might as well be. What the heck is that? She doesn't, you know, she can't do that. It's an exercise for me. It's impossible for a layperson. And I'm thinking, what why didn't you why didn't you the technology that I see is, you know, AR, VR helmets, VR goggles, and I'm gonna see my design come to life in front of me, or at least I'm gonna have a 3D uh fly through. I'm gonna see a 3D image. Please can have a 3D image, and I get drawings, and I think, what have you have you lived in a cave for 30 years? I did not know this technology.

Tudor 42:45

You're absolutely right. I think architects nowadays they owe their clients some sort of a tool, I mean, some sort of a way to make them understand what the heck they're doing. I mean, a floor plan is great.

Roopinder 42:58

Yeah.

Tudor 42:59

We can we can you know measure it and understand it fits in, or you can, you know, say a kitchen is that large, or you have an aisle that's that large. But yeah, like that. It's it's still um uh you know miles away from will I enjoy it in the end. That's where you know they could be doing their renderings in the least technical way uh type of software like SketchUp work. Whatever, but there is a big, big gap, and I've dealt with um architects that were like that, and slightly, I don't know, for some reason they were not uh pursuing 3D rendering in the studio, and they were reacting to early stage renderings as if, oh my god, this design is so bad, we have to change it. So you're on a time crunch, uh they're changing the design, their team is going wild and crazy, oh my god, it's too late to do that. No, no, no, it doesn't Work. It simply doesn't work. It's got to be improved. And everyone suffers. So I don't know what is the problem with digital.

Roopinder 44:12

It's there. It's there. It's surprisingly pervasive. And uh, I mean, it's not just me. Mine was a rather small project, but my another office neighbor has a two and a half million dollar residential project which involved um it's on the side. I don't know if you've ever been to San Francisco.

Speaker 5 44:30

Yeah.

Roopinder 44:31

Yeah? You across from San Francisco, there's uh salito and Tiburon, and a lot of the houses are built on very steep inclines. Like I call them cliff group dwellers. But uh but you yeah, you but you need the 3D there. A floor plan is not gonna help you because you need to see what it looks like in 3D. And yet, and I asked I asked him, I won't use his real name. Let's call him Jim. Jim says, I asked Jim, what did you show the client? How did you help them realize what you're doing? Because you're you're doing a two and a half million dollars, his biggest project ever that he's given anybody. How did you show your client this? I said, I'm sure they were impressed, right? And he says, I just showed them drawings. Oh no. I said, uh, and they said then he then he wonders how he didn't initially get the job. And and I'm thinking, boy, if you could have shown him the if he had used you guys, right? He said, you know, give me the images, I'm sure he would have dazzled the Yes, yes, and uh and but what I could say is that you know, as you have to be a responsible architect or designer, but you have to be a responsible client as well. You as the client are entitled to ask for you know something that is that is clarifying, that is a drawing is not enough. You've got to put in the effort to persuade, to convince me that what you're doing is is right. And you know, a house is it's way more than just you know a park, an outdoor park or a public building. It's some it's a place where you're gonna spend your life in. And uh these decisions, I mean, just building your own house, just but after seeing one rendering of the kitchen and one of the living room, it's still not enough nowadays. Right. It shouldn't be enough. Yeah. You need to see it. You shouldn't. You should be able to walk through, you should be interactive, you should be able to say, oh, what does it look like from here? Well, the how's the light come in? You should be able to live in the house that you're trying to make, or live in the office, or that you should be able to live there for a while or simulate it quite realistically. I think the expectation of technology is there, and the client, the client, you know, the client is actually pretty smart. Client has seen all the home improvement shows, they've seen the high-tech, but pervasive but predominantly the technology is not being used by architects, except, you know, except gratefully, you guys and people like you who are on the cutting edge, they can supply it. But but the long story short, you've got a big, you've got a lot of potential here. You've got there's a big market for people like that can do use the technology.

Tudor 47:22

I think I've and I'm not I I'm trying to see the um the good side. Yeah. Uh the good side is that this technology is gonna is gonna be uh being democratized. Yeah. And ourselves were trying to develop a product based on the on the platform that we build. Right. Um uh because we're seeing the potential, it's probably gonna take us out of business, or it's gonna move us into a different type of business. Yes. Um I can I can even demonstrate because you were you were saying, and video was doing this demo back in the day. Uh I can even demonstrate what what this tool does just with a uh like a screenshot out of uh out of SketchUp. I'll get to that in a minute.

Roopinder 48:04

Yeah. Do you want to do that now? You can put it in the show.

Who Should Use AI And How

Tudor 48:08

Oh yeah, we could we could we could um you know this is alpha. This is alpha uh uh alpha software. It's not it's not public yet. Yeah. But uh if you allow me, I'm gonna show you what NVIDIA's demo would look like today. Please. Um because I mean we're using this thing. Um this was this was it's called freestyler. Not sure if it's gonna stick or not.

Roopinder 48:39

Yeah.

Tudor 48:41

What what we're doing as our directors is we're trying to um so basically I can I can paste in and quickly show you through the process. I'm pasting here uh a screenshot, right? Right. And uh they've sent us um this screenshot. It's probably a basic uh analysis sketch of it with a bit of uh collage in there. Okay. And what what an art director does and what what a rendering person would do is let's say what I want from this image is to set the light, set the mood, um, set the time uh or the season or the sky, what the sky should look like. Do I want some weather like rain or snow or whatnot? Maybe fog snow or whatnot. Fog, why not? You know, being inspired by paintings or um beautiful artworking, but why couldn't I do that? And then maybe you want to touch up on the on the push the image up in the end with like some filters or like with some cinema style filters and you know put that into a dashboard. And what this thing is called, it's called an art rector because it really does what it exemplifies. This is a process put into the form of a dashboard. Okay. And instead of spending hundreds of hours just by testing these uh these assumptions here, uh we can just uh move the sun up in the sky, set its altitude, maybe I want some pitch black uh night or some sun coming up to decide, maybe let's get some uh rounders here, light fog. Um yeah, why not wet surfaces and weather maybe like rain. And just get better. Let's see what happened. Oh, and plus I did describe um what he's seeing here, so the the AI is hard enough to understand what's happening. I did describe uh this is a bright luxurious retreat resort um on a hilltop of a British British island. A climbing bright white white uh bright limestone and so forth, surrounded by green. That's the ocean there, framing should not change. And let's see. First ever live demo I've done.

Roopinder 51:20

Okay. Don't worry, well, if it doesn't work, we'll edit it out.

Tudor 51:24

It is. This feature here, but uh so it's cranky sometimes, it just places the sun in the opposite end of where you want it to be. But it's the uh, oh my god, and that's that's a fail. Okay. That's a massive fail. Uh it changed the view, and it just happens that you would probably get a different result. But you know, it's just me being brave and saying, you know, that happened.

Roopinder 51:53

I think it puts you in the middle of a swamp somehow.

Tudor 51:57

Oh, yeah, it does look like it's water and it moved the mountain now.

Roopinder 52:04

Yeah.

Tudor 52:04

But this one. Look at that.

Roopinder 52:08

That's not bad.

Tudor 52:10

Yeah.

Roopinder 52:11

Yeah.

Tudor 52:12

I think it added a tree somewhere, but it's just like the the team or any renderer would have spent a decent amount of time just scanning from you know from this to this.

Roopinder 52:28

Now, this is a tool that you've developed, Freestyler, and it uses Nvidia, and it uses you said you use Gemini for imagery as well.

Tudor 52:37

It does use Gemini, so it's uh built by Google. It's called Nano Banana.

Roopinder 52:42

Oh, sure, sure. Yeah.

Tudor 52:44

That's their top, that's their top model. And um it it just allows for you know very quick iterative work to be done um maybe you know, in instead of hours in seconds. In hours, if you're a very, very, very good um visualizer and fast and whatnot. But this thing is it's not using Nvidia, by the way. Okay. Um this runs in the cloud. It's um it's available to our team and to I don't know, some friends still. But um typically you could render these out in 4K resolution and be done with it. So if you're someone who gets like bad renderings from an architect that is designing your home and you want to see what they would look like with the proper material, so you want to change the materials on your home. Right. There's another blink, uh another glitch. It's probably the first time that it's getting created, but it it is that that that project. So it's making up what it shouldn't uh be doing, actually. It shouldn't change the view. It's taking a few liberties. It's changing. It's taking a lot of liberty. It's taking a lot of liberty, and it typically doesn't happen. But we're, you know, we're changing stuff every day, and that's why it's an it's an hour. It's just it's not a public software yet, design and data.

Roopinder 54:24

I might have missed this part, but Tudor, tell me, did you is the prompt, did you make the prompt, or did it did it create the prompt for you but from your slider bars and other controls?

Tudor 54:35

I made the prompt because I know what the client wants. I I know that they want this type of limestone and they need copper roofing. Okay. But this could have been done by uh you know having ChatGPT create your prompt for you. Okay. And it will be done in the app with potentially uh ChatGPT or Gemini either describing your image or improving this prompt to make it even better.

Roopinder 55:02

So it's a combination. It uses the prompt that you give it plus all the other uh slide controls that you've put into freestyle above that you're showing right here. Okay, got it.

Tudor 55:13

You basically can fool around with whatever parameters you want to create uh nighttime images and and whatnot that we could do right now, but still uh just spending no time uh to do anything except you know facing this uh screenshot and then waiting 15 seconds for it for it to produce something that's already gonna do a lot of uh you know good vibes or simulating the exact conditions um where the project is uh located.

Rising Client Expectations And Standards

Roopinder 55:51

Right. I it I see the sun direction, but are you able to give it uh time of day also and and location, geographic location? Yeah.

Tudor 56:01

Geographic location, not yet. I think that's um you would probably need to hint to where the north is in an image. So for now, um once you know where the north orientation is, you could be playing with this sun orb and placing it to the south, west, or east. Okay. Uh avoiding north. Okay. But in the future, I think what we can do is um we've got to we can tag each image, and that's a great idea, tag each each image with the orientation and have the software analyze and just place the light uh based on geographic orientation. Right. We can even use this simple uh configurator to say just what type of light you want and the blue hour uh or the midday or late afternoon, and it would uh automatically understand uh where the sun should be. Right, right. So you know you could experiment and um this is super cheap and you know and cost sense to generate an image. You can play around with it a lot. But I think that's where um I think the bridge between designing or design tools or big tools or whatnot and visualization would would uh rely on coordination between uh solutions like that and uh this sort of solution. So I would love to see BIM and AI sort of become responsible for the project and um uh you know helping architects do better for their clients. Because not seeing a rendering of your own home, that's not okay, in my view. Uh the architects should be at least literate enough. Even with this hand sketch, they sit up here to uh it could, uh you know, solutions like that they could help the architects express their ideas in a form that the client would understand.

Roopinder 58:08

Definitely. I totally see it as a sales tool. If you have this grand vision, uh architects are, you know, they're blessed in having an art uh artistic vision, but they're very limited in how they show it to the client generally. And I think that gap must be addressed because the full, let's say that the full genius is not on display in front of a client. And I think that needs to be that sense of space that a lot of architects have. I like to say this, I always admire another's what I don't have in myself, and I I can't do an artistic rendering, which architects are really good at having an image or or a vision of 3D space, we need to get that across the profession. And I think this is a great way to do it in two ways. It also helps people realize if we could show, if we could will help you show your portfolio, especially the latest project from in Naples, Florida, to show them what could happen on a large scale with a detailed project that's commercial. You know, they are, I'm sure they'll be very delighted that they can do a simple prompt and sketch up and get a rendered image, but in order to really do serious work, I think they need to know that there are services like yours that can help them. There needs to be an education here that you can't, you might have seen the you might have seen the commercial, but the real product is it takes a lot of uh know-how, process. Uh real architects.

Tudor 59:44

Probably, probably so. Um, customization is is a is a word that um keep keeps us grounded and keeps our texts uh grounded. And that's unacceptable. And that's a thing that that's a that's a that's a challenge for whoever creates um AI tool. How you you keep you maintain control of these parameters and the generation. I mean, this is just so early stage, everything that's related to AI. But um it is machine learning and it works and it it speaks machine in the end. Yeah. And you you work with parameters, and at the end of the day, everything will, with the help of uh AI researchers and machine learning researchers, I'm pretty sure uh tools like FreeStar or whatnot, all the other ones that are great already and available, um, they'll be able to offer architects and uh developers and ordinary you know, Myers, or homeowners. Uh it'll it'll be a it'll be an easy way to just test assumptions. Uh huh. Let's say you you want us you want to look at your living room and change the furniture nowadays with the AI. And then put the photo of the sofa you want to change, and it'll probably integrate it perfectly. You've got to watch out that for the dimensions though. Yeah. So is it gonna be accurate? Um I I don't I don't believe that that people would be obsolete uh or their services will be obsolete very, very soon, but on the long run, you have to watch out for what AI is gonna kind of change in the world.

Roopinder 1:01:33

It's I think it's like every shiny new object. I think people are gonna rush in, people are gonna try to use it, and and then people like us, engineers and architects, are gonna say, wait, it's not it's not accurate, it's not consistent, it's not it's not realistic, and and need to bring it back because we have to remember this tool was initially useful for students cheating on essay questions, and uh, and for that it's very important to have a different answer every time because you don't want your essay question to be exactly like the next student next to you, because for sure then you're gonna be cheating, caught cheating, right? So it was important for this in the beginning to have a different answer every time you questioned it. For us, we don't want that. We don't want the house to develop a new office or bedroom just because on the next because it has a whim to do so, right? No, no, keep exactly to be to be exact this dimension, but we don't want to change this specifically, change a roof, change limestone to to marble, change do that. Damn it, keep everything else the same, though. Don't yes.

Tudor 1:02:45

Absolutely. Absolutely. I I don't know how what what the state of AEC is at the moment. Um we have been waiting for Autodesk to implement stuff in 3ds Max, AI things. Hadn't happened yet. Um but they they do a lot of work with other tools uh that I know of. And so it's I mean they have a huge portfolio.

Live Demo: Freestyler Workflow

Roopinder 1:03:08

Yeah. They do, but you have a uh quick, fast mover, first mover advantage. Uh these are I call them sleeping giants. Uh they don't really appreciate that I do that. But you know, they're they're waiting to see, I think they're waiting for this to die out in some cases, AI. They're saying, okay, well, people will have their uh novelty uh AI for a while, but then they'll come back to us. I think that a lot of times they're waiting. They're implementing AI so slowly, and meanwhile, companies like I see are developing applications furiously. Uh and and I I think there's gonna be a there has to be a big change. All the big changes, not just Autodesk, but Autodesk, Dassault Systèmes, Siemens, all the uh Nemetschek companies, they're all very slow. Because you have to remember, they're mired in their process, they're mired in their code, right? They're mired in their code base, C, right? Are they gonna add Python to their technology software stack? Reluctantly, right? Maybe not. Their best effort in AI has been uh, okay, we'll help you uh with the documentation. You can decide how to do a command, and maybe it'll give you a good answer. And for us, it's like, no, no, you need to Okay, what am I getting to? This is like fast and agile startups have a big advantage right now before the sleeping giants wake up and discover that, hey, they better they better start implementing this in a in a more beneficial way.

Tudor 1:04:42

They don't I what what they what the what these um new companies are doing, they're potentially also using bytecode, which can get you uh results real quick, and if you're working with a with a responsible programmer or someone who's uh uh engineering, software engineering, you're luckier than others. But that that gives that that gives them and us uh a huge speed, and we can break stuff without and it's held by let's say it's held by Google. So you're you're you're not gonna be uh left, you know, just uh uh uh with your app uh down forever. You can reiterate, you can backtrack or you can re rewind. Um but it's it's allowing young companies to do what giants have taken years to produce. And what I think is gonna happen, BIM 2.0, these uh very light apps that can allow you to do uh feasibility studies, they will start having capabilities of of larger uh bit of software. At least up close. certain point. And it's not bad because I mean we are prisoners of of these of raising costs of software and it's it's so difficult as uh you know as any business to to sustain uh these costs of you know cost of living is rising a bunch of problems global politics and everything is just uh oh yes but but also the pressure from from these giants is increasing not decreasing and I do understand they must be they do have staff and they do have uh a lot of stuff going under uh you know behind the curtains uh these software is always evolve uh and evolve keep evolving but it's still not um you know the speed of AI is not uh gonna be uh stopping yeah no it's it's not it's not like they can they can keep up with it. Yeah. They can probably they if they need to rewrite their code they're done they will never do so. Right. It's gonna take years. Yeah. But they can uh glue a bit that is written in Python but it's um it's bridged by an API with their own software and they can be independent. And then you have an extension which is going to work in in there and uh it's gonna serve a different purpose or it's gonna inform your design and um you know maybe even vibe coded I'm sure.

Roopinder 1:07:32

Yeah yeah yeah you can do a lot with vibe coding and uh you're absolutely right I think there needs to be a a f another technology stacked onto what they have I never said that uh you should throw away your CAD program or you should throw away your simulation program or whatever, AC program. Those tools are really very good. They're very precise they're exact they're you know they're accurate they're dependable I just find them very difficult to use and I wanted to them to be easier to use and I want to be able to talk to them naturally like we're talking and give them prompts right I want them to understand what I'm doing and help me get there and not have to I don't want to be an AutoCAD architect I want to be an architect I don't want to be a Revit Architect. Yeah yeah I I think AI can do that. I mean for 30 years people have been asking me when I thought CAD was really advanced people have been asked asked me first every time I taught at a Autodesk University I thought I was on top of my game because I knew everything about AutoCAD and somebody in the audience asked me hey can I do a takeoff from this drawing? And uh and I had this I had to admit, oh no you can't do that right and now that AI came along I thought oh for sure I can do a takeoff now can I? You guys are going to give me a takeoff tool and still I can't do that. It's 30 years later and I'm thinking oh my God you finally have the potential to deliver on some of the things that you should have done in the first place right please let's not blow this opportunity let's use this technology and find a way to make it happen. Find a way to stack that technology find a way to you know deliver on what AI can promise in a realistic way. Don't just pass it off as hey I've had so many people tell me oh it's LLMs it's not designed for buildings or products or shapes right it's a language model and I'm thinking oh no that's not the point. It's a natural language interface to your product and you just have to realize that software can do more CAD could finally be computer aided design it it's never been computer aided. It's been computer aided documentation all this time it's never been aided design operated if you if you will because you have to operate inside there but it's never been I mean the user experience the user's experience can be fundamentally changed. Yeah you still need to to control your car if anything happens you probably need a steering wheel just for comfort yes yes not just for safety if something goes wrong that is um you have to be able to just uh set in your destination and probably just get worry about it. Exactly but the interface but just like in a car it's it's anybody can use it. It's easy to use I don't have to know how my engine works or how my microprocessor works or how it's figuring out how to get from point A to point B, but I need to be able to steer it in a natural interface. That to me is pretty natural a steering wheel I can jump from one car to another without having to read a manual or take a class I can do all those things because it's a natural sort of natural interface. I can't do that with CAD programs and why can't I? I can't jump to a visualization program because I have to learn it. And I'm thinking that's where AI could really come in handy and could really do some of those things.

Toward Smarter Light, Time, And Location

Tudor 1:11:15

So you have a lot of experience with uh AEC software and I've seen you know the you've done a breakdown of uh one of these um big 2.0 software on your website and I've I've also taken a deep dive into you know testing these um tools uh like Hektar and Giraffe and others and I have to say you don't have to know anything about CAD or designing or anything. I mean some are so easy to understand you just draw your your site on on the 3D map and then you hit generate and you you tell it I want a building that looks like that or you know has this sort of mass and you work with numbers better. You're gonna have to work with urban planning and stuff. But it you do get um reports in the end that would have taken I don't know so much time to produce in software like Revit which you should have learned before you've been able to do that. So I think the gap this gap is is being filled with uh people who understand that uh the experience of the user is more important you don't need to to learn uh so much no learning curve if you pick up the phone uh the smartphone you have to be able to use it without construction I think that's that's where the game should be for for most applications being built nowadays um if you're you're targeting you know uh non-specialist sort of people who shouldn't have to be put through all that uh that steep learning curve and I think it's happening yeah uh with with the help of AI yeah yeah it's but just uh these few these people understanding what the user needs in the end definitely definitely design and usability very true tutor it's been great having you on I don't want to keep you I think you're probably starving by now don't worry about it it's been a most informative discussion and I'm so delighted to talk to somebody because like I said you'll you always appreciate in others what you don't have in yourself and having a chance to talk to an architecture who has an artistic capability and talent is this is just wonderful and I'm so glad you can make that we'll be making that available to more more in the of our in our profession. Thank you so much for the these are I mean amazing words and I'm humbled um uh I I do hope that what we do here can be little as I said democratized and maybe you know I'm I'm I'm doing this for friends I'm doing this this for us for our uh studio but I have so many friends who are not fully illiterate but almost illiterate in rendering and creating imagery and they have beautiful portfolios but they never benefit with uh benefited from beautiful visuals and I know these guys are going to be helped by by this tool that is you know putting uh uh some artistic uh freedom in their hands uh at you know almost zero price um in the end but um I do appreciate your invitation uh what you do is fantastic I would I was humbled and scared to be among uh illustrious founders of amazing companies that I know of um so I recommend uh whoever's watching this to go to your YouTube and to your uh links uh to to watch these uh videos or listen to these interviews I mean there's so much it's a wealth of information down there for whoever's interested in what's happening in the universe of the AEC and not only. So um big big thanks and big hug um and keep up the amazing work I'm gonna keep following and um I'm learning very good great to talk to you have a good rest of your evening and I hope we see each other in person in the future.

Roopinder 1:15:34

Thank you for thank you for listening to the Future of Design and Engineering Software brought to you by ENGtechnica . I hope you have learned technology that will help you.