Artificial intelligence turns simple text into realistic building designs

Researchers develop a smarter image-generation system that produces realistic building designs with correct floor and facade details

Japan Advanced Institute of Science and Technology

Facebook

XLinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

The figure shows how the proposed framework turns a text description of a building into a realistic architectural image, step by step. First, the system uses the text prompt to generate a simple structural sketch that captures the overall shape of the building, including the correct number of floors. Next, the sketch is refined by adding detailed architectural elements, such as windows and doors, using a database of real building components as a reference. Finally, the refined sketch is combined with the original text description to produce a high-quality, realistic building image that matches the designer's intent.

view more 

Credit: Associate Professor Haoran Xie from the Japan Advanced Institute of Science and Technology

When working on projects, architects must quickly turn rough concepts into visual representations. Text-to-image models offer an opportunity in this field, where high-quality designs can be generated simply by typing a description. Some of these systems can also incorporate rough sketches or depth information, offering additional control over the results. However, these models often fail to generate accurate representations of the prompt. For example, even a direct prompt such as "generate a 5-story building" might result in an image of a building with the incorrect number of floors. The reason lies in the training datasets, which lack detailed annotations about building structure, making it difficult for artificial intelligence (AI) to understand precise spatial requirements, such as floor counts or the exact placement of windows and facade elements.

Researchers at the Japan Advanced Institute of Science and Technology (JAIST) have now addressed these problems with a retrieval-augmented generation system that combines text prompts with information retrieved from external architectural datasets, enabling the model to reference real architectural examples during generation. Such a tool could set the groundwork for AI-generated architectural design tools that make the process easier and faster.

The work, published online in the journal Frontiers of Architectural Research on March 26, 2026, was carried out by a collaborative team led by Associate Professor Haoran Xie from JAIST, together with Associate Professor Ye Zhang from Tianjin University, China.

“Today, high-quality architectural visualization requires significant expertise and expensive software. With the help of this work, individual designers and smaller teams will be able to participate meaningfully in the design of their own built environments, expressing preferences and seeing realistic results without needing a large professional team,” said Dr. Xie.

The team designed the framework to mirror real architectural practice. Architects typically begin with simple sketches that show the overall shape and layout of a building. Over time, these sketches are gradually refined with more detailed elements, such as windows, doors, and facade components. The new system follows this step-by-step process.

First, the system converts the text prompt into a simple structural sketch that captures the overall building form and ensures the correct number of floors. Next, it refines this sketch by adding detailed architectural elements using a database of real building components. Finally, the refined sketch is combined with the original text description to produce a realistic, high-quality building rendering that accurately reflects the designer’s intent.

To evaluate the framework, the researchers tested it on campus building designs, where controlling the number of floors and the placement of windows and entrances is especially important.

They constructed three specialized datasets: a building box dataset containing 2,200 images, a component dataset with 4,000 images showing different window and entrance arrangements, and a sketch–rendering pair dataset with 1,600 examples linking detailed sketches, text prompts, and final campus building renderings.

In objective evaluations, the framework achieved 70.5% accuracy in vertical configuration and outperformed baseline diffusion models on several quality metrics measuring structural accuracy, visual realism, and alignment between generated images and text prompts.

The results were further supported by a subjective study involving 56 graduate students in architecture and design. Using a five-point Likert scale, where 1 indicated “very dissatisfied” and 5 indicated “very satisfied,” participants gave the system average scores above 4 for image quality, alignment with prompts, and architectural detail accuracy.

Such a system could significantly improve early-stage architectural design workflows. “Designers can use it to quickly revise schemes in response to client feedback during meetings, dramatically shortening the design iteration cycle. Planners and developers can use the tool to visualize and compare dozens of design alternatives under shared constraints before any detailed modeling begins,” explained Dr. Xie.

As AI continues to evolve, tools like this could make architectural visualization quicker, more accessible, and more reliable.

  

This figure demonstrates how the proposed framework was applied to a real campus building design project. The left column shows the actual site plan and photographs of the completed building, provided by Tianjin University (reproduced with permission). The middle column shows building facade designs generated by a standard artificial intelligence (AI) model without our retrieval-augmented approach, while the right column shows designs produced by our framework. By comparing the two sets of results, it is clear that our framework can generate building designs that better match the specific style and constraints of a real campus environment, more closely reflecting what an architect would intend to build.

Credit

Associate Professor Haoran Xie from the Japan Advanced Institute of Science and Technology Image source link: https://www.sciencedirect.com/science/article/pii/S2095263526000452?via%3Dihub

***

 

Reference
DOI: 10.1016/j.foar.2026.01.018

Authors: Zhengyang Wang, Yuxiao Ren, Hao Jin, Jieli Feng, Xusheng Du, Ye Zhang, and Haoran Xie

 

About Japan Advanced Institute of Science and Technology, Japan
Founded in 1990 in Ishikawa Prefecture, the Japan Advanced Institute of Science and Technology (JAIST) was the first independent national graduate university to have its own campus in Japan. Now, after 30 years of steady progress, JAIST has become one of Japan’s top-ranking universities. JAIST strives to foster capable leaders with a state-of-the-art education system where diversity is key; about 40% of its alumni are international students. The university has a unique style of graduate education based on a carefully designed coursework-oriented curriculum to ensure that its students have a solid foundation on which to carry out cutting-edge research. JAIST also works closely with both local and overseas communities by promoting industry–academia collaborative research.  

Website: https://www.jaist.ac.jp/english/

 

About Associate Professor Haoran Xie from the Japan Advanced Institute of Science and Technology, Japan
Dr. Haoran Xie is an Associate Professor at the Japan Advanced Institute of Science and Technology (JAIST) and Waseda University, where he directs the Human-Centered AI Laboratory. His research focuses on human-centered generative AI, exploring how emerging technologies can enhance human capabilities through interactive computer graphics, deep learning, and human-computer interaction. Dr. Xie's work spans creative applications, including anime, architecture, and fashion design generation, as well as physical intelligence and robotic learning.

 

Funding information
This work was supported by the JST BOOST Program, Japan, Project Number JPMJBY24D6, and the National Natural Science Foundation of China, Grant Number 52508023.

---

Journal

Frontiers of Architectural Research

DOI

10.1016/j.foar.2026.01.018 

Method of Research

Computational simulation/modeling

Subject of Research

Not applicable

Article Title

Controllable Generation of Building Representations: Aligning Campus Building Design Intent with Multi-Stage Retrieval-Augmented Diffusion Models

Article Publication Date

26-Mar-2026

COI Statement

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

 

Phylogenetically diverse central China proposed as newest global biodiversity hotspot

Chinese Academy of Sciences Headquarters

Facebook

XLinkedInWeChatBlueskyMessageWhatsAppEmail

Taxonomic endemism and phylogenetic endemism are both important measures of biodiversity. The former describes the number of distinct species found nowhere else, whereas the latter shows the amount of evolutionary branch length unique to a particular area. A comprehensive phylogeny provides the essential evolutionary framework for delineating centers of paleo- and neo-endemism across both measures.

Historically, global biodiversity hotspots are defined by exceptional taxonomic endemism among plants, accompanied by severe habitat loss: regions must host at least 1,500 endemic vascular plant species and retain less than 30% of their original natural vegetation. To date, 36 such hotspots have been recognized worldwide.

Recently, a research team led by Prof. LU Limin and Prof. CHEN Zhiduan from the Institute of Botany of the Chinese Academy of Sciences, along with collaborators from Australia and the United Kingdom, identified a new global biodiversity hotspot in Central China, based on traditional measures of taxonomic endemism.

At the same time, the researchers also examined patterns of phylogenetic endemism across China—offering new insights for optimizing biodiversity conservation and supporting implementation of the Kunming–Montreal Global Biodiversity Framework.

The study was published in Nature Ecology & Evolution.

The researchers first reconstructed the most comprehensive dated phylogeny of China's vascular plants, encompassing 3,029 genera (99%) and 16,585 species (53%) native to the country. By integrating over 1.4 million distribution records with this phylogeny, they revealed a mismatch between taxonomic and phylogenetic endemism centers.

The researchers showed that taxonomic endemism centers are concentrated in southwest China's Hengduan Mountains, in Central China, and in the Yunnan–Guizhou–Guangxi boundary region. In contrast, phylogenetic endemism centers extend further into northern China, including the Tianshan–Altai Mountains and the Changbai Mountains.

Notably, the researchers identified Central China, an area of approximately 1.54 million km2, as an important area for global biodiversity. They noted that this region supports over 14,000 vascular plant species and also serves as a key center for insect and vertebrate diversity. Despite retaining only about 7% of its original vegetation, it harbors at least 2,024 endemic vascular plant species, meeting the criteria for global biodiversity hotspot designation.

All in all, Central China's subtropical evergreen, broad-leaved forests represent a unique confluence of ancient relict lineages (plants that were once widespread but are now found in only a few regions) and recent rapid radiations (many new species have emerged over a relatively short period).

Based on these data, the researchers proposed formally designating Central China as a global biodiversity hotspot—to safeguard its diversity and irreplaceable evolutionary heritage from accelerating anthropogenic threats.

By focusing on both taxonomic and phylogenetic endemism, these findings provide a blueprint for future conservation planning, ensuring that protected area networks capture both species richness and deep evolutionary history. If recognized as a global biodiversity hotspot, Central China would become China's fifth such hotspot, thus strengthening its role in global conservation efforts and attracting increased international support.

---

Journal

Nature Ecology & Evolution

DOI

10.1038/s41559-026-03025-1 

Article Title

A comprehensive dated phylogeny of China’s vascular plants reveals a hidden global biodiversity hotspot

Article Publication Date

31-Mar-2026

 

Cow manure digesters really cut methane — unless they leak

Study finds rare failures can have enormous consequences

University of California - Riverside

Facebook

XLinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

Methane plume over a cow manure lagoon in California on June 22, 2023.

view more 

Credit: Alyssa Valdez/Google/UCR

A new study shows that systems designed to capture methane from cow manure, called dairy digesters, are highly effective. But on the rare occasions they fail, the leaks are large enough to offset their climate benefits.

“I think manure emissions on dairies are underestimated. These digesters seem to be a solution that captures a lot of methane,” said Alyssa Valdez, a University of California, Riverside climate scientist and lead study author. “But I wanted to make sure they were working properly.”

The findings of her study, published in Environmental Research Letters, draw on eight years of satellite and airborne observations of 98 dairies across California. By tracking emissions before, during, and after digester installation, Valdez and her research team were able to see how these systems perform over time and at scale.

Digesters are widely seen as a key climate solution. By sealing manure ponds and capturing the gas they produce, these systems convert methane into usable fuel instead of allowing it to escape into the atmosphere where it has a tremendous effect on the climate.

Methane is shorter lived than carbon dioxide, but it is 80 times more powerful at trapping heat in the atmosphere, making even small releases significant.

A previous study led by UCR climate scientist Francesca Hopkins examined emissions at a single dairy using ground-based measurements. Hopkins found that a well-managed digester can cut methane emissions by as much as 80%. This new research builds on that work by showing how digesters perform across dozens of farms, including what happens when things go wrong.

Across the dairies studied, the number of strong methane plumes declined after digesters were installed, suggesting the systems are effective overall. However, the researchers also detected occasional leaks that were far more intense than emissions from traditional manure storage.

“For the most part, the digesters are working well,” Valdez said. “But the few leaks that happen, they make a huge impact.”

In some cases, the team observed methane escaping at rates around 1,000 kilograms per hour. By comparison, typical emissions from open manure lagoons ranged from 20 to 100 kilograms per hour.

The contrast highlights a central challenge: digesters concentrate methane in one place, making it easier to capture, but they also increase the risk of powerful releases if something goes wrong.

Those large releases are not limited to system failures. The study also captured spikes in emissions during digester construction and installation, a phase that is rarely measured but can produce substantial short-term increases.

To capture these patterns, the researchers relied on satellite and aircraft data. Satellite images allowed them to track changes across dozens of dairies over long periods, which is not possible with traditional ground-based monitoring. Aircraft measurements were then used to identify concentrated methane plumes over specific infrastructure locations, making the approach especially useful for spotting leaks.

“A farmer might not know their digester is leaking,” Valdez said. “This gives us a way to detect issues early and prevent them from becoming long-term problems.”

However, this method does not capture all emissions. It cannot measure more diffuse methane releases from sources such as lagoons or fields. For that reason, the researchers say satellite and airborne observations are most effective when combined with on-the-ground measurements, which provide a fuller picture.

This need for comprehensive monitoring comes as California continues to invest in digesters as part of its strategy to reduce emissions of heat-trapping gases. Hundreds of these systems are already operating or in development across the state.

In some cases, methane releases are not accidental. Operators may vent gas when it cannot be flared due to air quality regulations or when systems require maintenance. These process-related emissions add another layer of complexity to managing digester performance.

Even so, the study shows that most systems are working well and that large leaks are relatively uncommon. But for Valdez, who spent years living in California’s Central Valley, and whose family lives there, the work is about ensuring that climate solutions deliver real benefits in a region critical to the nation’s food system.

“This region is the backbone of our food supply, but people there also carry a lot of fear about air quality,” she said. “And they have good reasons for that.”

More broadly, the study highlights the need to pay closer attention to agricultural waste.

“We need to start caring about poop,” Valdez said. “And we need to keep verifying that these solutions are actually working. If we monitor them carefully, we can make sure they deliver on their promise.”

---

Journal

Environmental Research Letters

DOI

10.1088/1748-9326/ae4fe4 

Article Title

Evaluating the impact of anaerobic digesters on point source methane emissions from California dairies from remote sensing

Article Publication Date

24-Mar-2026

 

Towards smarter agriculture: Durable nanofilm electrodes for monitoring leaf health

Researchers develop non-invasive devices for the long-term measurement of bioelectric potentials in plants

Institute of Science Tokyo

Facebook

XLinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

This is the first transparent, durable, and water-resistant electrode that does not compromise the natural functions of leaf hairs. It could enable real-time monitoring of plant health status and help secure higher crop yields.

view more 

Credit: Institute of Science Tokyo

Nanofilm electrodes capable of detecting stress in plants through bioelectric potentials could pave the way for more resilient agriculture, report researchers from Institute of Science Tokyo. Thanks to the electrode’s small thickness, leaf surface hairs can easily pierce through it, enabling stable and long-term electrical contact without compromising the leaf’s natural processes. This work could help improve crop yields by enabling early detection of stress in plants.

Food security is becoming a major global concern as climate change and growing pesticide resistance put crops under pressure. One promising strategy to tackle these issues and improve yields is catching health problems in crops early, before they cause irreversible damage. Interestingly, much like animals, plants produce electrical signals when under stress, and measuring these bioelectric potentials directly from leaves can provide timely warnings of diseases or harmful conditions. If developed into practical sensors, such technology could become a key component of smart agriculture systems.

However, despite many research efforts over the past decade, existing leaf electrodes are not yet suitable for widespread adoption. Some electrodes are not transparent, which interferes with photosynthesis. Others are not fully water-resistant, making them unsuitable for long-term use. One particularly tricky obstacle is the presence of trichomes—the tiny hair-like structures found on the leaves of many economically important crops, including soybeans, tomatoes, and eggplants. Existing thin-film electrodes are designed to simply cover trichomes, disrupting their function and damaging the leaf over time.

To address these issues, a research team led by Professor Toshinori Fujie, graduate student Yusuke Hori, and Assistant Professor Tatsuhiro Horii from the Department of Life Science and Technology, School of Life Science and Technology, Institute of Science Tokyo (Science Tokyo), Japan, in collaboration with Professor Shinji Masuda from the same institute, developed a new type of durable thin-film electrode compatible with trichome-rich leaves. Their latest paper, published online in Advanced Science on March 23, 2026, describes transparent nanofilm electrodes made from conductive, single-walled carbon nanotubes deposited onto a flexible elastomer layer. Only 70 to 320 nanometers thick, the nanofilms are extremely thin and can conform closely to complex leaf surfaces without the need for adhesives.

One key discovery was that the 70-nm-thick electrodes could be pierced by trichomes, settling directly onto the leaf’s epidermis while leaving the hairs largely undisturbed. This “trichome-piercing” mechanism—consistent across several crop species—enabled stable electrical contact without damaging plant tissues or compromising key biological processes. The electrodes also proved highly transparent, transmitting over 80% of incoming light and allowing photosynthesis to continue normally.

Tests showed that the nanofilm electrodes could record bioelectric signals continuously for several weeks. In some experiments, the devices remained attached and functional for up to 10 months without causing apparent damage. The research team also confirmed the electrodes’ durability in simulated rainfall conditions; unlike hydrogel-based sensors, which fail after exposure to water, the proposed carbon nanotube electrodes remained attached and continued to record electrical signals without issue.   

The researchers also showcased how their electrodes could be used to monitor physiological stress in plants. “When the electrode was attached to leaves under herbicide damage, chemical stress was successfully detected through changes in the bioelectric potential waveform associated with light irradiation,” explains Fujie.

Because electrical signals often change before visible symptoms appear, the proposed devices could enable early detection of plant stress in real-world settings. “Our findings make it possible to non-destructively capture physiological changes that occur before stress levels reach the stage that leads to yield reduction,” remarks Fujie. “In the future, we expect this technology to be applied for crop health monitoring in agricultural fields.”

By enabling long-term monitoring of plant electrical activity, these new nanofilm electrodes could become a valuable tool for precision agriculture. With further development, networks of these sensors could help farmers respond more quickly to environmental stressors, marking an important step toward more resilient food production systems.

 

***

 

About Institute of Science Tokyo (Science Tokyo)
Institute of Science Tokyo (Science Tokyo) was established on October 1, 2024, following the merger between Tokyo Medical and Dental University (TMDU) and Tokyo Institute of Technology (Tokyo Tech), with the mission of “Advancing science and human wellbeing to create value for and with society.”

---

Journal

Advanced Science

DOI

10.1002/advs.202522824 

Method of Research

Experimental study

Subject of Research

Not applicable

Article Title

Pierceable, Water-Resistant, and Transparent Nanofilm Electrodes Comprising Carbon Nanotubes for Long-Term Monitoring of Plant Electrophysiology

Article Publication Date

23-Mar-2026

COI Statement

This work relates to the patent to be pended: JP2024-131682.

 

Climate change could make unhealthy air a routine reality by 2100

New modelling shows a dramatic rise in air quality alerts, with vulnerable populations facing the greatest risks

University of Waterloo

Facebook

XLinkedInWeChatBlueskyMessageWhatsAppEmail

New modelling shows almost one in three Americans will routinely breathe air considered unhealthy for sensitive people by the year 2100 due to climate change, a sevenfold increase compared to the turn of the century. 

The international study, led by the University of Waterloo, found that about 100 million people in the United States will live in areas where average air quality during smog season is poor enough to trigger alerts advising vulnerable people to stay indoors. 

That is up from an estimated 14 million people in 2000, with most of the increase coming in California and the eastern United States. Smog season runs from the beginning of May to the end of September. 

“Climate change could cause days with poor air quality to shift from rare to commonplace,” said Dr. Rebecca Saari, a professor of civil and environmental engineering and the Canada Research Chair in Global Change, Atmosphere and Health at Waterloo. 

“People who are especially sensitive to air pollution, including the elderly, children with asthma and those with health conditions, could face a daily coin flip, with nearly even odds of an alert every day asking them to change their behaviour to reduce exposure.” 

The study built on previous research that estimated the number of air quality alerts in the  United States will quadruple, and that staying indoors to avoid the health risks posed by worsening air pollution due to climate change would require an additional 142 days per year by the end of the century. 

The new study broadens the picture by including the impact of both ozone and particulate matter pollution, which together cause almost all air quality alerts and are the primary environmental contributors to sickness and premature death. 

The research team also examined the economic costs of intensifying air pollution and the potential mitigating impacts of policies to limit it over the next 75 years. 

Saari said a significant finding is that seniors benefit much, much more than other vulnerable people from steps to limit their exposure to poor air, such as staying indoors, wearing masks, and improved building filtration. 

“We were surprised by how widespread and common it could be for sensitive groups to experience air that is considered unhealthy on average,” said Saari. “We hope our work helps inform more targeted guidance and reinforces the need for pollution reductions and long-term adaptations such as access to clean indoor spaces.” 

Collaborators included researchers at Harvard University, North Carolina State University and the University of California, Davis. 

Their study, Air Quality Alerts, Health Impacts and Adaptation Implications Under Varying Climate Policy, appears in Environmental Science and Technology. 

---

Journal

Environmental Science & Technology

DOI

10.1021/acs.est.5c12522