Introduction

Okay, good afternoon. My name is Luis Candeias.

I'm the founder of Experimental Space, how you see, and I'm talking here about imagining our future in architecture with AI.

Well, for start, I'm an architect, and I started this one year ago, developed some AI. I work for two big companies. One is from AI, another is OpenAI, the founder of ChatGPT.

And I will show you how was my, in this year, how I work.

Social Media Impact

So the first thing is that I do some image from my work in architecture. I went in Instagram very fast in one year from 18K.

I have 2.3K posts with AI, some experiments from different companies. I don't use only two companies, I use five companies to do image for them.

And it's very, very, very important for me, for my work, to develop some architecture, not only exterior but interiors, and how this architecture could be a must in the world of AI.

Well, for me AI, artificial intelligence in architecture is create innovative architectural solution that enrich communities, inspire creativity and the human experience. This is what for me is the artificial intelligence that could good have solutions for the architecture.

AI in Architecture

For do some image in AI, in architecture, we need a very good prompt. What is a good prompt?

A prompt is nothing... better than something like this.

1AI is transforming architectural design by helping architects optimize their creation. It's a machine learning algorithm that can analyze vast data sets of building performance, environment factors, and user preference to suggest design modification, improve energy, efficiency, structure integrity, and overall functionality.

This is not only a plus for sustainable buildings, but also reduce construction costs. We are talking about

much further than AI can do in architecture. Well, this is some example of what architecture could be not only in sustainable but and structural efficiency, whatever we need in architecture.

This is very important for us architects because we have some ideas and we have to transform these ideas in concept. And this is very good AI for some ideas that we have and we don't have the skills or the design or the projects.

And we have AI for a very big tool, and I only have this tool for giving some ideas. This don't build, this don't do projects, this is only idea and conceptualization.

Well,

Creating Effective Prompts

It all starts with simple phrases like unique architecture from in the forest. What is this? This is something that we can show to the clients what we have concept in your mind and what the clients could have about

perspective on how we can change the the architecture in the environment and the integration in all aspects of the surroundings this could be another an amazing exhibition central retrofuturist another client that could help and ai could say that we have different forms different shapes different lights different surroundings this help us to show to the client what we are trying to say this is very very fundamental in the world of ai

So, and the prompt that we use is normally some what we have in our mind. This is very simple, what you see there, but it's more complex what you see.

Well, I start the prompt like a stunning aerial photo of a touristy cityscape showing advanced architecture with water, canals and lush gardens integrated into design. What I was saying here is that I want a mix between architecture, buildings, and nature.

And this is what AI gave to me, like the last for image. But this is more than this.

Prompt Structure

The prompt is normally divided in three main characters. We have the craft, that is text prompt, image prompt. text, image prompt, then we have the exploration, generating image, diversity or depth of prompt, and the result.

We can upscale image or the final image. What I can say, the upscale image is the level of detail that we want to show to the clients.

The craft is very important. The first thing that we do is a text prompt, only using text. We can change the text, we can adapt, and we can see what the image could change with just some words.

The image prompt is we get some image from reference, and we adapt this image to see that, imagine that one building that is from Mies van der Rohe and we adapt this to a modern or futuristic architecture. Text and image prompt, it's a combined text image that we can allow AI to say what we want and generate something, the mix between the two.

The exploration is the generation of the image. Diversity is the multiplication of the image and adapt the prompt. If we don't like something, we just adapt the prompt and it changes immediately the image.

The final image we can use to show the clients. And we have the prompt structure.

The prompt structure, it begins like the views, then the architecture, the place, the seasons, the materials, the environment, the colors, the others. It's not just so simple like the first sentence that I show you in the beginning.

It's like cinematic drone views. It's the views of mesmerizing photo captures, the amazing futuristic form design architecture, what we want.

Like an equipment, it is a museum in the middle of the park. It will be the place in the north of Ireland.

We will be the season, like autumn, water materials. We are just describe the materials, the environment, and then the colors, what we want, and other things that we want to add.

The image is created. We can change, for example, some sentence in this, and the AI adapt what we are seeing, what we are... describing, imagine that you don't want in the Northern Ireland, or you don't want a museum, or you don't want futuristic farms, you want parametric farms, organic farms, and you can change. And all the AI adapts from this thing.

Examples and Collaborations

This is the image that I produced for Dolly Tree. It's a company of OpenAI.

Some of them that I use to change prompts, and the image can be like this only the prompt never change i have this perspective is only view architecture place season material environment colors and others others could be people could be cars could be other things that you want to change

This is Ideogram, another company that I have collaboration and it's the same prompt and you can see the changes. The same prompt is nothing like this. It's always this prompt, but you can see the change.

Dolly Tree, Ideogram. and now from AI.

This is very different. The result, this is why I test a lot of programs in AI, because they give us different image and different environments and different architecture.

I start to integrate this architecture with surroundings, with the environment. And I think it's very good, the work that AI do in this field.

Because for us, like architects, it's very hard to integrate the things with the surrounding, with the nature. in the city, even that we can do this, but we can show to the client these ideas. It's only ideas and conceptualizations, okay?

Challenges and Adaptations

Well, this is the hard stuff. It's not so easy like this.

And I have this. To process is generate a lot of image. Testing prompts, change prompts, changing image, change text image and I always say that 30 is the number minimal image that we could do for a best good final image.

We are always changing, we are always mutating the forms, the views and you can see more than 30 that I have to see to do this good image that I want to show to the client. Then is the diversity in the architecture that AI could help us

and change or put the to be the plus for our diversity not only in sustainability but all the results this is about nature and my conceptualization was in the middle of the nature some cabin in the woods but different cabin woods and different conceptualization this is just simple image, I can change the prompt, I can change some words, but the final result is always almost the same.

We can change the lights, we can change the surroundings, but it's not like a thing that is very easy to do. We can change environments, we can change the buildings, we can change the aspect, what we want,

And this is what in the beginning when I started AI one year ago, I started to do some interiors and some exteriors, low light, the surroundings was very dark and then I changed the dimension. This is a museum and the other is an auditorium, a parametric auditorium that is going to be a lot of light it was for one client and I start to change to do some change in this architecture well now nowadays I do more sculpture and I do more

other kind of equipment, but I'm more artist, art, and then architecture. This just was the last image. Inspired by nature or materials, we can add, like I said in the beginning, some image of materials, and we can do a plus to some architecture.

You can see the bubbles. We can transform buildings in bubbles. We can change some flowers in one building too.

this is the aspect that is going we can make a tree or from the hook or we can put bone and convert in structure this is only parametric buildings we can change all what we adapt without with image and prompt and text This is a net and this is a bug inspired by season two. We can inspire by season.

Integration with Nature

This is inspired by the surroundings. We can adapt the buildings to the seasons, to the places, to the environment. and inspire what we want.

I start to play with some expectancies, the flowers, the flower shop that I start to do to a client and it's very great. The result is like inspiration and ethereal. The other is Mashmallow with the possibility of structure and implementation of futuristic design and lights.

We can have some coffee shops in the middle of the nature. We can integrate these coffee shops in the middle of nature and not only in the environment but also in the middle of some urban space. Urban space, now we can integrate the nature in the buildings.

In the middle of urban space, we can put only one building in the nature and is different the aspect. We can play with materials. This is cotton steel, buildings made by cotton steel and different, the same prompt, different aspect, different diversity and other places.

We can change this to urban, to night. This is always, always thinking and what AI could transform these buildings and could what we have in mind and how AI transform this, what we have in our imagination. It's a plus.

No, I don't have any doubt. This is the surroundings about materials, integration of woods, integration of different aspects of surroundings. And this is more complex.

It's parametric. And I have some tests here about animals, flying birds. And it's amazing.

The results of AI is spectacular. Even me, I keep learning from AI a lot. This is an auditorium for a client that I did.

And it was at night. The client was saying that they want a lot of light, like a a lighthouse, and the result was amazing.

This is a house, a simple house, modern and in the middle of a cave or in the middle of the rocks. Some expect that we do some scaling. This is auditorium, the materials are using for nature.

We can say to AI that we want some materials from nature and it converts in some facades or some buildings or some structures that we want to show to the clients. We go to the urban space and the urban space is very, very hard to make some image of the reality because I don't like the image like 3D. I prefer the image that are very real and we give some aspect different to the client.

This is two buildings in streets in New York and one is parametric, the other is geomorphic. And now we can say that we can transform image in video, like four seconds, and we can change all this with only AI.

Future of AI in Architecture

machine learning is not about building a perfect system but one that is better today than it was yesterday.

I hope you connect with me in the Instagram is experimentalspace and if you have any ask question or something I'm glad to help you. and I'm here to do some practice now.

I will create some image for you, okay? And I will use two programs. One of the programs that I use is from AI.

It's pay. It's not free. I'm sorry.

But you have a lot of programs in the field that are free. For example, Ideagram, Leonardo AI. I work with them too. And it's amazing.

so we will use a simple prompt okay and i will show you how no matter what the prompt is only small texas like you see there and we can change the text we can change all this prompt wherever you need i will do some copy paste just to see how is the result You have a lot of information in these apps. Prom.ai is one of them, for me, is one of the best. Like Dolly Tree, too, is the app that we'll create.

You can have some style, scene, perspective, proportion, illumination, and if you go this, I will show very quickly, okay? The illumination, the proportion, what you want, the perspective is what you need. the scene is whatever you decide, and the style, okay?

So I will use only this for default. I will use V2 because it's the last render, if I can say, the last render or the last implementation of AI. I will show you what this do. We will generate three image, okay? With the prompt, it was in the beginning.

I hope it's first. And then we will do the same prompt with Dolly Tree. Dallytree was created by OpenAI about one year ago. I started with them with this implementation of Dallytree. It was very good work what they do. They are trying to develop an app and I don't know how it's going to be. It will be an app like this, from AI, it will be pay or not.

Comparing Different AI Tools

Well, this is the simple. prompt that we do, then we can adapt, we can change all what we want, no matter what, surroundings, materials, persons, seasons. Now we will do the same prompt with Dolly Tree. And you will see the change. We'll put it again.

You can see the changes are very different from one AI to another AI. The concept, the construction, the data are very different. This is more artist. I like very much what is the concept of Dolly Tree.

Because when I do the prompt, I start to, like, when I want to do more cinematic things, I double the words, like cinematic things, a dreamy atmosphere, amazing colors, aesthetic. Aesthetic is a very strong word for what we want, like theatrical buildings or something. And you can see the difference.

But we can... We have more. One I will try for you.

It's Ideogram. Now it's another company, a very big company in the AI for architecture, not only for architecture. This AI could do design, could do a lot of things, not only architecture. And you can see here.

And I will put the same prompt. You can define landscape, square, but we will do some image. And you will see the difference between three AI.

Well, from my experience from one year ago, I experiment like 10, AI, apps, and all are different. The results are always different.

What means is not only is the receiving data that they will receive, but they are not similar. We can say, oh, the Dolly Tree is the same thing that Prometheus. No, it's not. It's very different, and the images are very different from one part, more conceptual, more integrated with nature. We will see another difference with another. Meanwhile, this is done another one.

I will do another one. We can change the prompt. Now I change the material. Not carton steel, but wood. It's the same prompt, but only the material are going to change. I will do the same in Promei.

Meanwhile, Ideogram is doing. And you will see the difference between the three. This is the change for wood, not cotton steel. And then I will try one more that is with paper, like the building were like paper, OK?

This is the result from the with wood. Just change the aspect of the materials, OK? And we will do. with paper, like the buildings be like paper.

We will see the results. You can see the change from wood to paper. And you can apply wherever you want.

The limit is our imagination. More conceptual, more artists, different kind of buildings, different surroundings, different aspects.

and another from PROM.AI, different construction, different forms. And now we will see if ideogram, yes. Well, Corten Steel is the building, the same PROM, but different aspect.

You can see the difference between what is ideogram and what is PROM.AI. I will show again with cotton steel and with dolly free. Okay, so this is the same dolly tree from AI and ideogram.

They are very different. They are not even similar, not only in architecture, but the surroundings, even the aesthetic thing, the material are more accurate in some programs than in others, but well.

Conclusion

we have to try and this is very good for architects because we can show to the clients what we have in our imagination it's more fast yes it is it's more accurate yes and it's a plus for architecture not only what i said to you in the beginning but i will keep you this in mind That is, AI is transforming architectural design, helping architects. I don't have no doubts on this.

Why? This optimizes their creation. Yes, it's optimized.

But I just want to keep you in mind, this is only a tool. a tool for create ideas or create studies. The project has to be done with engineers, with the normal data set, planning and other things.

This is only, only the beginning of what it could be a building or small coffee or other things. Well, end with this. Thank you very much for seeing my presentation, and I hope you like it.