§ work · No. 031
Research by Models
2022–2024 · researchPhysical models, left unfinished, augmented by image models
- year
- 2022–2024
- type
- research
- credit
- Daniel Koehler. Models in mixed media; augmentations generated with diffusion models, 2023. Published through exhibitions and as figures in Compositional Intelligence.
- program
- Synthetic Types
- no.
- 031
The series tests the concept of inductive types with objects rather than prompts. Each model exposes its structural frame, its geological base and its architectural elements in different states of completion; a diffusion model is then asked to read the object and to propose what it could become. Because the model responds to what is physically there, the augmentation stays answerable to the thing on the table.
The method produced figures for Compositional Intelligence and the studio briefs of Zero Mile Assemblies, and it taught me the rule I have followed since: a generated image is only as good as the physical thing it has to answer to.
plates
Gallery
14 plates · click to enlarge
The models, augmented · generated images, 2023 
Contact sheet of generated model photographs · 2023 to 2024 
A model made of windows and terraces, standing in a studio, 1 · generated image, Midjourney · 2023 
A model made of windows and terraces, standing in a studio, 2 · generated image, Midjourney · 2023 
An architecture sculpture of arid elements, 1 · generated image, Midjourney · 2023 
An architecture sculpture of arid elements, 2 · generated image, Midjourney · 2023 
An architecture sculpture of dry elements, 1 · generated image, Midjourney · 2023 
A 3D-printed apartment model in a white studio · generated image, Midjourney · 2023 
An X-ray styled tower-house model in a white office, 1 · generated image, Midjourney, upscaled · 2023 
An X-ray styled tower-house model in a white office, 2 · generated image, Midjourney, upscaled · 2023 
An X-ray styled tower-house model in a white office, 3 · generated image, Midjourney, upscaled · 2023 
An X-ray styled tower-house model in a white office, 4 · generated image, Midjourney, upscaled · 2023 
An X-ray styled tower-house model in a white office, 5 · generated image, Midjourney, upscaled · 2023 
An X-ray styled tower-house model in a white office, 6 · generated image, Midjourney, upscaled · 2023
1 / 14 · The models, augmented · generated images, 2023
adjoining sheets
Related
3 entries
Compositional Intelligence
A monograph on generative AI and architectural typology, with a foreword by Mario Carpo. How large models compress the built world into types, and how architects can compose with those compressions rather than consume their images.
Designing Cities with Cities
Generated image series from 2022–23, made from and about the imagery of cities: compressed city types, remixed neighborhoods, treehouses, urban material studies, and Rasa Navasaityte's interiors. The pilot studies that led to the LCA-GCS dataset.
The Gießhof
A projective design study for Stuttgart: urban soil read as ore, iron cast on site, clay rammed into piers, and a cooperative that owns the result. Six grounds, one section, six institutions; developed alongside the studio Synthetic Types, 2026.
§ work
More work
34 entries
Complicit Code
A drawing shown in Drawing Codes: instead of beginning with setbacks and height limits, the buildings are designed through consent and complicity governed by blockchain algorithms. Mashburn Gallery, Houston, and Richardson Memorial Hall, New Orleans, 2025.
Architecture After AI
The first overview of generative AI in architecture: more than 2,000 works by 80 architects and artists, shown at the Mebane Gallery in Austin in 2023 and twice in Boulder in 2024 and 2025. Curated with Kory Bieg and Clay Odom.
Foresting Architecture
Can cities be built with living trees? An advanced design studio at UT Austin and an image series that explore an architecture literally forested with trees; the series won the Metaverse Architecture and Design Award 2023.
Carbon Hinterlands
What about Cows: the future of parking and other carbon hinterlands. An image series on the parking garages that autonomous vehicles will leave behind, returned to farms, gardens and pasture, shown in Solarpunk Futures at P61, Berlin, 2023.
Large City Architecture Research
Twelve years of research from five institutions in one room: dozens of white models on a scaffold of cardboard beams, set against the carbon arithmetic of how Austin builds. Curated by Daniel Koehler and Rasa Navasaityte, 19 January to 15 April 2022.
Light Architecture
Planting with UV light: a study of the spatial synergies that the artificial lighting of urban indoor farming brings into buildings, in which the light needed for vertical planting figures the voids of new urban interiors. Published as a chapter in Diffusions in Architecture (Wiley, 2023).
Living with Trees
Rasa Navasaityte's contribution to Potenziale 3 at aut. architektur und tirol, Innsbruck, 2021 to 2022: a housing proposal that simulates the interaction of building parts to study the in-between of human and natural voids, shown as models, screens and a wall print in the Digital Lab InterActs.
Potenziale 3
The lab's two contributions to Potenziale 3: Insights into Architectural Research at aut. architektur und tirol, Innsbruck: the models and drawings of Rasa Navasaityte's studio Living with Trees and Daniel Koehler's video installation One Minute Architecture, with the large print made for the show.
Three-Nine
Three architecture machines arranged 382,659 building parts into variations on a nine-square grid, resembling the daily spatial use of 500 people in downtown Austin. A cloud-based consensus simulation developed with Unity Technologies; the model was built for the Mebane Gallery exhibition of 2022.