April 15, 2024

how and when to use generative design

Cas Esbach

In the changing landscape of architectural design, artificial intelligence (AI) is not just a trend, but marks a significant shift in practice. Early adopters have already explored various AI tools and software. Now is the time for the design community to embrace this change. And there are specific and practical ways that AI can refine and improve the design process.

As we navigate this revolution, the role of AI goes beyond mere efficiency. It’s about enriching the creative process, improving communication, and opening up new possibilities for complex and nuanced design solutions.

So how can you integrate AI into your architectural design processes and potentially go beyond the pioneers by forging new paths in creativity and innovation?

1. Concept development with AI

The effectiveness of AI stands out in the early stages of concept development. After the initial brainstorming and sketching phase, where various ideas are presented, integrating a tool like Midjourney can be a strategic power move.

Midjourney will quickly generate images that closely align with each concept outlined. Sharing these AI-created images within the team helps clarify and solidify everyone’s understanding, moving beyond the sometimes vague nature of some sketches while also demonstrating the validity of others.

This clarity ensures that everyone is aligned and discussing the same vision, and the same approach is invaluable when communicating with clients. By replacing abstract sketches with clear AI-generated images, conversations with customers can become more focused and direct. This change in the way we present and discuss ideas marks a significant improvement in communication, making the conceptual development process more efficient and precise.

Source: Cas Esbach

Message: Sami cultural center in Finland, wooden structure, attractive, welcoming, crowds gathered, misty, –ar 3:2. Time to create 10 minutes.

2. Generate ‘references’ – create your own precedents

Searching for the right references to illustrate design ambitions can be time-consuming, often involving hours on Google or Pinterest. Despite the effort, it is often difficult to find the perfect reference image.

This is where Midjourney comes into play. With its ability to quickly generate custom images, we can create images that not only fit our needs, but are also unique and not associated with other companies’ work. This capability enhances our ability to communicate our design intentions clearly and effectively.

Source: Sandra Baggerman (MVRDV)

From a creative office with lush plants to flowers to an office consumed by plants, anything is possible. Creation time: 15 minutes.

3. Modeling 2.0

Modeling is a crucial part of the design process. This often involves carving foam blocks to roughly represent the project within its intended location. During brainstorming sessions, the team breaks down these models and reorganizes them into new designs. This stage is a chaotic and creative part of the design process, where unexpected twists and turns can lead to the most innovative solutions. Sometimes even a lost coffee cup or water bottle can become part of the table setting.

With the introduction of AI, we can bring these foam models to life, vividly demonstrating the potential of each design direction. This technological advance allows us to visualize our ideas in the real world with greater clarity and detail.

Source: Cas Esbach

From blue foam model to an initial impression in minutes showing scale, materialization, atmosphere and the way you will try to deal with the vegetation. Creation time: 20 minutes.

4. Enhanced Mass Studies: AI as a Design Catalyst

As we progress through the design process, our preferred modeling technique changes. The initial foam blocks are replaced with abstract 3D models created in Rhino or SketchUp. These models allow us to calculate gross floor areas (GFA), test core designs, reconfigure surrounding roads, etc.

While these models are informative, they are often very abstract. Simple massive screenshots struggle to communicate the full potential of a design choice. This is where AI comes in, transforming massive screenshots into captivating images that make both the team and clients fall in love with the project. The use of AI-generated images effectively bridges the gap between an abstract 3D model and a realistic representation without requiring a large investment of time.

Source: Cas Esbach

From an abstract mass to a more captivating design using Stable Diffusion Time to create: 20 minutes

5. Details and advanced studies of materials.

Traditionally, materialization studies involve slow and careful development of facades, testing panel dimensions, corner solutions and expansion thicknesses for a wide range of materials. While this is an inherently creative process, the great effort required to develop each direction often limits the opportunity for spontaneous and innovative ideas.

However, by incorporating Stable Diffusion (a text-to-image deep learning model released in 2022) with a more developed 3D model, we can quickly explore hundreds of design possibilities. Since the AI ​​is basically making educated guesses in its renderings, it reintroduces an element of fun and surprise into this phase of the design, just like foam models. This method allows for a more dynamic and less laborious exploration of materialization options, which can help limit the design directions that can be further developed.

Source: Cas Esbach

Artistic facade + glass railings; golden + ornamental facade; stucco facade + integrated plates; brick facade + New York balconies; Wooden facade + white steel railings. Pink tile facade. Creation time: 60 minutes.

6. Display communication

Stable Diffusion is not a complete substitute for a professional visualization studio. But it does serve as an excellent communication tool between clients, architects and visualization teams. Typically our decision making is based on sketches rather than raw renders with poor basic lighting and texture mapping, which can lead to regrettable decisions once the final images are revealed. By integrating AI, we can quickly produce and evaluate various views and lighting configurations.

These AI-generated images become valuable assets in our conversations with the visualization studio, significantly reducing the likelihood of miscommunication. It’s an approach that streamlines the decision-making process, ensuring that final renderings align more closely with the initial vision envisioned by the team.

Source: Cas Esbach

A rainy night, a warm autumn sunset, a snowy winter day or cherry blossoms blooming in spring. Creation time: 45 minutes.

7. Final thoughts: opening new doors

AI is not just a tool, but a transformative force in the field of architecture. From the initial spark of conceptual development to the fine details of materials studies, AI is capable of improving, clarifying and innovating. Each step, whether creating striking images with Midjourney or refining design communications with Stable Diffusion, demonstrates the role of AI in not only simplifying but also enriching the architectural process.

As we embrace these advancements, we open doors to new possibilities, where the limits of creativity and efficiency are continually redefined.

This integration of AI into the architectural design process marks a new era of design, one that balances technological advances with the artistic and intuitive spirit of architectural creation. By effectively leveraging these AI tools, we move towards a future where our designs will not only be more efficient and accurate, but also more in tune with the changing needs and aspirations of our society.

Cas Esbach is project leader at MVRDV, based in Rotterdam

Leave a Reply

Your email address will not be published. Required fields are marked *