AI in ArchViz
There’s a new kid on the block in the ArchViz world, and it goes by the name of generative AI.
This technology has sent ripples through our industry, like many others.
It has opened up new possibilities and solutions for creating and editing architectural visualisations.
But it has also raised some questions and concerns among some 3D artists and designers
Will generative AI replace us?
Will we lose our creativity and control?
Will we become obsolete?
Our answer to that is no, no, and no.
Generative AI is not a threat, but an opportunity.
It’s not a replacement, but a complement.
It’s not a limitation, but an enhancement.
But it is a tool that requires us to learn and adapt, or get left behind.
We need to become creative explorers, able to experiment and discover new ideas and scenarios as part of a new and exciting symbiotic working relationship with generative AI.
It’s early days, but I want to share with you our initial experience with generative AI and how it has changed our workflows for the better.
I’ve been creating and teaching 3D imagery for over a decade using SketchUp and V-Ray.
But when I tried generative AI, I was amazed by what it did for my workflow and output...
What is generative AI and how does it work?
Generative AI is a technology that can generate images from text descriptions using algorithms and machine learning.
It can help you create stunning all kinds of visual media, including architectural visualisations, in minutes, with minimal effort and maximum flexibility.
But how does it work?
Let’s find out.
How generative AI works
Generative AI works by using two types of neural networks: a generator and a discriminator.
A neural network is a type of machine-learning model that mimics the structure and function of the human brain.
It consists of layers of interconnected nodes that process information and learn from data.
A generator is a neural network that creates new data from existing data.
Generators can create new content from the likes of images, video, music, speech, text, and software code inputs at the time of writing. A typical text prompt may look something like this:
'A modern living room with a fireplace, a sofa, and a coffee table'
A discriminator is a neural network that evaluates the quality of the generated data.
It compares the generated image with real images and tries to determine if it is fake or not, and scores it from a scale of 0 to 1.
The generator and the discriminator work together in a feedback loop, where the generator tries to fool the discriminator, and the discriminator tries to catch the generator.
The generator learns from its mistakes and improves over time until it can produce realistic images that can fool the discriminator.
This process is called generative adversarial learning, and it is the core technique behind generative AI.
What are some generative AI tools for ArchViz?
Generative AI has many applications in various fields and industries, but below I want to focus on some of the tools that are relevant for the ArchViz, Architecture and Interior Design industries.
These tools can help you with different aspects of your work, such as generating ideas, creating concepts, producing images or videos, or animating scenes.
Here are just a few examples of generative AI tools for ArchViz:
DALL-E: A generative AI tool that can create images from text descriptions. It can generate realistic or surreal images of anything you can imagine. You can use it to create concept art, mood boards, or illustrations for your ArchViz projects. Most of the imagery in this blog was created using Bing Chat which is integrated with DALL-E 2.5
Adobe Firefly’s Generative fill: A generative AI feature in Photoshop that lets you create and edit images with simple text prompts. It can generate, add, remove, or replace images with just a few words and a click of a button. You can use it to create stunning ArchViz images in minutes, with minimal effort and maximum flexibility. More on this later.
Canvas: A tool by NVIDIA that can generate realistic landscapes from simple sketches. You can use it to create backgrounds or environments for your scenes.
Midjourney: Another generative AI tool that can create dream-like images from text prompts, very similar to DALL-E. It can generate artistic and abstract images that reflect your vision. You can use it to create all manner of imagery to use for your ArchViz.
Stable Diffusion: A tool by OpenAI that can animate static images and make them come to life. You can use it to create dynamic and interactive scenes from your images.
These are just some of the examples of generative AI tools for ArchViz. There are many more tools out there that you can explore and experiment with, with tons more in development.
I want to focus on AI SketchUp extensions in an upcoming blog/webinar, so stay tuned!
What are the benefits and challenges of generative AI for ArchViz?
Generative AI offers many benefits for ArchViz artists and architects, such as:
You're never starting from zero: The ability to create concepts from plain language prompts is immensely powerful and means that you and your clients are on the same page from day 1.
Faster design development: At the time of writing, generative AI can help you save time and money for your projects by automating some of the tasks that you have to do manually, such as correcting things in Photoshop, filling in details and textures, enhancing the quality and realism of an image, editing and modifying an image according to feedback, etc.
Enhanced client experience: Generative AI can help you improve your client satisfaction and loyalty by offering more value and service to your clients, such as generating ideas and inspiration for their projects, producing images or videos that match their style or vision, animating their scenes and making them more dynamic and interactive, etc.
However, generative AI also poses some challenges for ArchViz artists and architects, such as:
Accuracy and quality issues: Generative AI can create artefacts that are inaccurate or low-quality. These could include blurry, distorted, or unrealistic parts of the image. You have to be careful and vigilant when using generative AI tools and always validate and verify their outputs or feedback.
Ethical and legal issues: When handing over the reins to generative AI, it could well create imagery that is offensive, harmful, or deceptive. You have to be responsible and respectful when using generative AI tools and always follow the ethical and legal guidelines and regulations of your field and industry.
Identity and creativity issues: Thinking back to how generative AI works - it ingests information and precedents and then uses this to create your desired output. As a result, you need to be aware of plagiarism and copyright infringements, rather than blindly accepting what is being produced.
How AI can help you communicate and collaborate with your clients more effectively and efficiently
What if we could ask for some assistance in the early concept stages with our client? Let's see what happens if we run the following prompt through Bing Chat (powered by DALL-E 2.5):
I want to design a new home for my client. They love the style of Frank Lloyd Wright's Fallingwater. They have a flat plot of land - 30m x 30m. Here are a list of their requirements; 2x floors, driveway, adjoining garage, office in the garden, large terraced area for entertaining, build the house into the land and landscape with natural grasses and drought-resistant vegetation, large, modern oak front doors
Interestingly, it also advises the following:
That sounds like a very exciting project. You have a clear vision of what you want your client’s home to look like, and you have chosen a great inspiration in Frank Lloyd Wright’s Fallingwater. I have created some images of how the house might look like based on your description. You can see them below.
To design the house in more detail, you might want to use some home design software. There are many options available online, some of which are free or offer trial versions.
Whilst the suggestions are not perfect - when designing, we no longer need to start from zero, and we can use generative AI to get us started.
Quick, intuitive edits
Let's look at a simple example of how it can help with communication and collaboration with project stakeholders more effectively and efficiently.
We were provided with the following backdrop to add in to our 3D scene
We wanted to show more of the sea, make the clocktower more pronounced and remove the overhead cables & streetlight...
Enter Adobe Firefly...
In this context, we used Adobe Firefly to mask the areas that we wanted to remove, and it took care of the rest.
It understood the image that I had uploaded and judged what it should show instead in those masked areas that I wanted to remove.
This is just a small example that demonstrates the power of generative fill for situations where you need to make edits to existing imagery for the purpose of creating ArchViz.
How to adapt and evolve with generative AI for ArchViz?
Whilst generative AI presents a time for change, we see this as an opportunity for ArchViz artists and architects. It is not a substitute but a complement to your work.
It is not a competitor but a partner for your creativity.
To make the most of this opportunity, you have to adapt and evolve with generative AI.
You have to learn new skills and use new tools. You have to find new ways to add value and differentiate yourself from the competition.
You have to embrace generative AI as a game-changer for ArchViz and architecture.
Here are some tips on how to adapt and evolve with generative AI for ArchViz:
Education is still fundamental: You still need to have a solid foundation of design, composition, lighting, colour, perspective, and other fundamental concepts for creating quality work. Generative AI can help you with some of these aspects but it cannot replace your knowledge and intuition.
Keep updated with AI developments: You need to keep up with the latest developments and trends in generative AI and understand how they affect your workflow. You need to be aware of the possibilities and limitations of generative AI and know how to use it effectively and ethically.
Communication is still essential: You need to communicate with generative AI tools and collaborate with them. You need to learn how to write prompts or commands for generative AI tools and how to interpret their outputs or feedback. You need to learn how to guide and control generative AI tools and how to make the final decisions and adjustments for your work.
Co-creation is still possible: You can co-create with generative AI tools and leverage their power for your benefit. We choose to see them as a partner rather than a competitor for your creativity.
Become prompt engineers.
There is no denying that generative AI is a technology that can revolutionise your archviz workflow and output.
The only creative limit you will encounter is your ability to articulate your designs.
In practical terms, we mortals are still required to gatekeep the content produced by our intelligent and creative companions.
This a technology that moves fast and requires you to learn and adapt, or get left behind.
You need to become creative explorers, able to harness the power of generative AI and discover new possibilities and solutions for your ArchViz projects.
In this post, we've barely scratched the surface of what AI can do in this industry, and this is something that we will be dedicating more time to in the coming months to ensuring that you stay ahead of the curve.
There have been huge advances in 3D to AI-generated renderings in recent times, with several SketchUp extensions being released, with more on the horizon.
We have plans in the works to create more informative content like this to help you, as SketchUp users, navigate this new frontier of generative AI in archviz.
In the meantime, if you or your studio need some tailored consultancy or training sessions, reach out to us in the chat below. We would love to hear from you.
We are certified V-Ray for SketchUp artists and specialise in offering archviz training, where we have taught thousands how to create photorealistic internal and external imagery.
We have a self-paced, Complete Guide to V-Ray for SketchUp which will help you create stunning renderings of your 3D models, with realistic lighting, materials, and composition good practices.
If you want a heads-up whenever we release new content like this - sign up here.
Thank you for reading and happy creating!