Back to articles

ARCHITECTURAL RENDERS IN THE AGE OF AI

The rise of electronic tools and AI in creating photorealistic architectural renders has prompted a question: Is there a place for traditional hand-drawn sketches, or do these high-tech renders risk creating unrealistic client expectations? We delved into this issue by speaking to two innovative architects, Andrew Kudless of Matsys and Harshdeep Arora (prspctivs), who employ state-of-the-art software programs to navigate this new frontier of AI.

KEEPING IT REAL

 

First up: Harshdeep Arora aka prspctivs, an architect from India, who has worked in Delhi, Shanghai and Beijing for renowned firms like UNStudio, Buro Ole Scheeren, Plasma Studio, DADA Partners and AECOM.  He’s currently working as a part-time senior architect consultant at CHAP (Cook Haffner Architecture Platform), while successfully pursuing freelance projects as an AI Artist/Architect. He’s making quite some waves with his awe-inspiring designs.

 

It all started back in 2021-2022: the launch of the initial models DALL-E and Stable Diffusion sparked Arora’s curiosity in AI. The introduction of Midjourney, and the “wave of surreal images” that followed, sealed the deal. “It was at this point that I felt a strong desire to explore these processes myself, especially because I had witnessed countless hours wasted on the manual process of 3D drafting and photorealistic rendering over the years.” The learning curve has been steep at times, mainly due to the lack of professional tutorials. Arora: “To overcome this challenge, I delved into reading numerous notes, referring to YouTube discussions by fellow AI enthusiasts, and, most importantly, engaging in hands-on practice to devise my own processes for learning and improvement.”

 

“Embracing AI in the creative process opens up new possibilities and allows designers to push boundaries and realize their visions in ways that were previously unimaginable.”

 

Harshdeep Arora

PAYING IT FORWARD

 

In the meantime, Arora has made it his mission to inform others about the possibilities and limitations of AI-based tools. “I aspire to motivate people to embrace this revolution and adapt themselves to create more than what they were previously capable of or trained for. Embracing AI in the creative process opens up new possibilities and allows designers to push boundaries and realize their visions in ways that were previously unimaginable.” Although that doesn’t mean computers are ‘taking over’: “AI has made great strides, but precise model generation using AI is still in its nascent stage. As a result, architects continue to rely on traditional 3D modeling and rendering for final designs that will be physically implemented on site.”

 

Speaking of physicality. Many of the designs Arora creates will most likely never materialize and he’s okay with that. It comes with the job: “Over the years, working on concepts and designs of various scales, I've noticed that a significant percentage of large-scale projects never materialize. Many architects find themselves involved in design competitions or commissioned projects that may get stalled midway due to various reasons, such as client budget issues or government interventions. While this might seem like a waste of effort, this process is essential in creating the exceptional 10% of high-quality buildings that do get realized. It's through this iterative and exploratory approach that architects push the boundaries of design and create remarkable structures.”

OPEN SKETCHING & CHEAP IMAGES

 

Andrew Kudless, trained architect, professor of architecture and head of Matsys design studio, is one of those pushing the proverbial envelope. “I sketch with AI in a similar way that artists and architects have used hand sketches to dream possible futures. Sometimes I have a specific goal but more often, it is a form of exploration to help me expand my imagination,” Kudless explains. “I often describe the feeling of using generative AI tools as a form of open sketching. Although I’m not using a pen or pencil to imagine the work, it has the same give-and-take that sketching has.”

 

“These images range from the photorealistic to the painterly to the surreal. However, all of the images exhibit impossible elements, and this is what draws me to them."

 

Mark Kudless

"The AI model does not understand the rules of light, geometry, or structure, but it does a reasonably good job of simulating them. And in this simulation, there are glitches that help me think about my work in new ways.”

 

Mark Kudless

 

Kudless also points out that despite looking almost identical, generative AI images are in some ways the exact opposite of traditional digital renderings: “Renderings are expensive in that they require immense time, labor, and skill. They provide the designer with highly accurate simulations from multiple views. AI images are cheap in that they take nearly no time or skill and will never provide a completely coherent simulation of reality, especially from multiple views. However, there is great value in this cheapness. I can create thousands of variations on a theme in the same time it would take me to create one finished digital rendering of one of my designs. Used correctly and in partnership with traditional design modeling and rendering workflows, AI images can support a designer’s early explorations.”

Veils play a prominent role in the work of Kudless. And when looking at the AI generated works he has put out, it’s tempting to think that the veils are one way of showing what rendering programs can do with transparency and layers. But that is not the case. “I have been fascinated with using fabric in architecture for most of my career. It started in 1998 when I was living in Japan. You can see one of my first projects in 1998 documenting the hundreds of fabric-covered construction scaffolding in the Kansai region,” Kudless explains.

 

"I think it is important to test and use a variety of models as they have been trained with different datasets and provide different approaches and functionality. The AI space is evolving so quickly that you need to be open and curious.”

 

Andrew Kudless

 

When asked if he has ever had any concerns about living up to the picture-perfect renders he presents to clients, Kudless replies: “As a designer, I always try to make sure the built reality offers something more than could ever be captured in an image. The image is an invitation to a much deeper experience. I used to say that I always wanted to make things that were more wonderful in reality than they were in renderings. And despite the advances in AI images and rendering technologies, I still think this is true as an image will never be able to capture the full spectrum of visual as well as non-visual sensory stimulations we experience in reality.” At the same time, Kudless still thinks it’s important to create beautiful images for they are symbols of the aspirations for a project: “For the designer, they can serve as a guidepost to keep us on track as there are always so many issues in the design process that can lead us astray. And for the client and community, they can rally support in ways that a plan drawing cannot.”

image002-1-scaled.jpg

We're very proud to call NOOK one of our partners. 

NOOK is the official magazine of the Dutch and Belgian professional associations for interior architecture, BNI and AINB. The magazine offers in-depth content, inspiration, and opinions. NOOK is published quarterly in English. 

The above article was first published in NOOK, the Drawing issue (2023 / #3). 

Comments

    Sign in to reply and see more comment(s)