Harshdeep Arora

Advisor: Shawn Bailey

Diagram of architectural designs with sketches and text overlays. Title: "Diffusing Design: Exploring Architecture through Latent Diffusion & Machine Learning." Modern aesthetic.
Architectural design collage featuring labeled sketches and photos of office interiors. Images highlight ceilings, workstations, and storage. The tone is modern and innovative.
Thesis Panel
Diagram illustrating Stable Diffusion through ComfyUI, showing interconnected modules with text prompts, settings, and nine resulting 3D building images.
Stable Diffusion Workflow using ComfyUI node-based interface.
Collage of architectural photos, with left side showing 15 varied building styles labeled "Training Images," and right side showing 6 labeled "Diffusion Results."
Diffusion Model Training using Site Surroundings (left) showing potential results (right)
Modern building with vertical fins and angular forms next to a digital model. Both feature geometric lines. The scene is dynamic and futuristic.
The generative diffusion interpreted digitally, showing various aspects like façade, materials, massing, scale, and placement on site.
A split image showing a library interior. Left side is a sketch with rough outlines, and the right side is a realistic digital rendering, bright and modern.
Adding intent and control to generative diffusion process through sketches.
Grid layout showing modern architectural design processes. Each of the eight rows has an interior rendering, a digital model, and a floor plan update.
Diffused design selection, digital model interpretation and sequential plan updates for first and second floors of the planned facility.
Complex architectural design concept illustrating window and wall integration. Includes diagrams, materials, and detailed annotations with a focus on light control.
Façade development through generative diffusion & conversations with LLM (Large Language Model).
Architectural elevation drawing of a modern building with large glass windows and sleek lines. Elements detail materials, window-wall ratio, and structural components.
Design summary showing collection of technical and design considerations selected and applied to create this iteration of North façade.
Modern architecture featuring a gray, industrial building with large windows and sharp angles. Two people walk along a path with dry grass under a blue sky.
An exterior collage, overlaying developed digital model and various generated elements through context trained models.
Architectural model of a modern building on a wooden base. It features clean lines, large vertical slats, geometric shapes, and bare branches.
Physical model of an iteration at the 2nd Quadrant Design.

DIFFUSING DESIGN    
Developing Architecture Design through Latent Diffusion & Machine Learning

Studio Description:
Machine learning and diffusion have created a new path for creation. Achieving tasks through machine learnt systems, especially creative work for designers, opens new possibilities.  But the base data of diffusion trained on existing architecture & context questions the credibility & possibility of finding something New. The methodology to create through diffusion process follows alternative unconventional trajectories. Also, developing architecture through diffusion has its challenges, to be identified, addressed & possibly solved.

As diffusion process adds Communication involving language/text/speech & visual/image referencing, in addition to conventional methods of Click & Command, conveying and processing ideal results involves various steps. Just like verbal & visual conversation, communicating intent in diffusion is a two-way process requiring input/guidance, assessment, editing & re-communication.

Hence, this research presents various processes/workflows in diffusion resembling/inspired by contemporary means of developing architecture, achieving a variety of designs. Exploring workflows in guiding diffusion, experimenting through Sketching, Referencing, Language, these workflows & experiments will make way towards enabling designers to guide, communicate & learn with creative machine learning, critiquing upon its use, limitations, and future possibilities in Architectural practice.  

But what can we achieve by using Diffusion? Experimenting methodologies inspired by conventional architectural means like sun/shadow study, massing, sketching, this research tries to iteratively design contextualized spaces guided by constraints & parameters. Designing with Diffusion not only can create intent, but more, creating anomalies, artifacts inspirations unfamiliar to a Design to Draft means, creating opportunities to find and develop with the diffusion process. Combining these conventional techniques with diffusion, including adding site, context & culture opens unique opportunities for design, vision and sensitizing this process specific to its use case. Communicating and finding in this hybrid design process takes new paths employing convention & generative diffusion. Like a music conductor, my role as an architect in this study will be guiding these experiments, adding & refining necessary conventions (programs, circulation, & others) creating wholesome architecture.

The promise of diffusion models creating a variety of ideas and results may also highlight their potential drawbacks. The designs generated through these diffusions, proposed for a specific purpose, function, and location, must be rectified and optimized. Hence, an architect must apply their own knowledge in the form of site sensitivity, material choice, climate conditions, programmatic requirements and surrounding connections. Detailing at both macro and micro levels, exterior and interior, material and technical, through professional expertise, while simultaneously involving generative AI for iterations and suggestions, whether for façade optimization, interior layouts, lighting, and other elements, will be explored throughout this study, with a collection of Iterations, Tests & subsequent Logs.