The contribution examines the case study of the Church of S. Pietro a Coppito in L’Aquila, which underwent significant stylistic restoration in the 1960s, to explore the potential applications and reliability of generative AI in cultural heritage communication. In particular, according to the virtual reconstruction of the ancient configuration, the study investigates the impact of integrating graphical inputs alongside textual prompts for image generation, assessing their effectiveness as a means of guiding AI and improving the quality of the results achieved. Based on graphic, photographic, and historiographical sources, sketches of the interior views of the church were produced. Subsequently, various gen-AI platforms were tested to process these images, evaluating the algorithms’ ability not only to generate coherent and effective visuals but, above all, to understand and interpret the initial graphic inputs. The results show both potential and critical limitations: while AI can produce photorealistic and stylistically coherent renderings, it often introduces arbitrary elements and struggles with strict historical accuracy, especially in architectural and decorative details. The paper highlights the need for deeper human-AI interaction and emphasizes the importance of informed graphical input and critical evaluation in the application of genAI to cultural heritage representation.
Artistry, Technology and Challenges: The Subtle Balance Between Fake Results and Real Integration in the Use of AI for Image Generation on Medieval Frescos Reconstructions
AI-based image-generative tools significantly enhance creativity and digital reconstruction by accelerating workflows for reconstructing digital heritage. These tools apply to various contexts, such as archaeological sites, transformed urban areas, damaged buildings, unbuilt architecture, and partially lost artworks like mural paintings and frescoes. In reconstructing such works, the process is highly complex and requires scholars with expertise in recognizing and attributing fragments and better-preserved pieces. While it remains necessary to suggest potential lines, figures, or scenes, the outcome often involves multiple possibilities. AI image generators can support these challenging tasks by integrating them into workflows, from producing stochastic results to extending patterns and coloured areas. However, their use, especially via “off-the-shelf” software, introduces two key challenges. The first concerns achieving accurate reconstructions that maximize the AI’s potential while avoiding casual or tentative outputs. This involves addressing limitations in existing AI systems and ensuring the final results are both reliable and time efficient. The second challenge relates to cultural evaluation. The line between AI-assisted reconstruction and creating entirely fake results is delicate. Misuse could exploit public fascination with AI or lead to errors and misinterpretations without expert oversight. This integration is in its early stages, necessitating rigorous testing and exploration. The proposed contribution analyzes the digital reconstruction of medieval frescoes from central Italy, blending traditional methods with significant AI inputs. These case studies highlight the balance between human-driven processes and the acceptance of digital outputs, offering reflections on the evolving relationship between human expertise and AI-driven creativity.
The Language of Photography in the Age of AI
Text-to-image algorithms based on Deep Learning are central to content creation in multiple application domains. In the last few years, the capacity of Neural Networks to generate increasingly realistic images quickly has blurred the boundary between authentic and realistic content, making genuine and false data less and less distinguishable. This condition leads to a profound reflection on the application of photographic images as a tool for communication and storytelling, trying to answer simple questions. Can today’s Neural Networks generate content comparable and indistinguishable from a photograph in both formal and compositional terms? Can artificial intelligence algorithms replace the photographer’s ability to design and obtain images that preserve the story and the place’s intangible culture? From a set of photographic rules framed in specific workflows, the research analyses some results obtained using text-to-image algorithms within the Midjourney program. The experiment aims to determine the pros and cons of using text-to-image algorithms to automatically generate photographic images, highlighting the potential and current limitations in constructing content subject to specific formal rules.
An Educational Experience Between AI and Architectural Drawing
This paper presents the latest phase of a research project investigating the interplay between the representation of ‘Virtual Living’ and digital technocultures—particularly the integration of Artificial Intelligence (AI) and Extended Reality (XR)—within the educational framework of an Architectural Drawing course. This course is part of the third-year curriculum of the Bachelor’s Degree in Architecture at the ‘G. d’Annunzio’ University of Chieti-Pescara. The research builds upon and refines technocultural methodologies long employed in teaching, leveraging the concept of the ‘semantic model’ as a versatile foundation for designing habitable virtual spaces, such as metaverses or virtual museums. This approach has been revisited and expanded to address the rapid evolution of generative AI applications, which demand rigorous monitoring and continuous thematic experimentation. New technologies in representation are reshaping the pedagogical landscape, offering unprecedented opportunities to redefine the scope of architectural drawing education. From descriptive geometry to surveying, from the history of representation to design, the incorporation of AI has fundamentally transformed how visual representations are conceived and executed. This paper discusses a case study that bridges research and pedagogy, showcasing how students’ creativity, when coupled with the capabilities of AI, facilitates the creation of innovative semantic models. These models have direct applications in the design of Virtual Cities and Museums, offering a vision of inhabitable spaces within the metaverse.
Txt2city. From the Prompt to the City’s Image
In the urban sphere, AI is often associated with the concept of smart cities and thus the use of technology and the enormous computing power of machines to increase the quality of life for citizens, creating greater efficiency in resources and services, but there are further applications. The sprawl of information technologies is changing relationships between people and space, contributing to the mutation of iconographic production and the way content is conceived and communicated. Digital management and communication processes focus on images. The set of images constitutes a highly evocative language; it can be immediately comprehensible or require decoding that refers to specific contexts and cultures. Their importance is evident in human-targeted communication, but they also constitute a data transmission vehicle for empirical knowledge generation by algorithms like those used for artificial image creation from other images. Text-to-image AI generators are trained through the analysis of hundreds of millions of images and their related textual descriptions, which allows the system to learn the relationship between text and visual elements. Through this process, the network is also able to infer other information about reality. Images can be the representation of actual objects, as well as a subjective product of imagination or sensory processing. So too are the images created by Italo Calvino’s Invisible Cities, which bear witness to mental and non-geographical spaces. Cities that cannot be seen can be constructed from their poetical descriptions. What textual variables affect the realization of an image? How do the datasets that different text-to-image tools draw on affect image realization? Can these tools be trained from the user’s realization of an image? The paper collects the first outcomes of a research project comparing the results produced by the main image-generation tools from text descriptions extracted from Calvino’s book.
GPTfor Treatise Image Creation: ACritical Overview
Text-to-image systems based on generative pre-trained transformers have become pervasive in recent years. This result is mainly due to the generation tools’ ease of use, application availability, and increasingly refined graphical and algorithmic capabilities. On the other hand, the non-controllability given by a random process and the dependency on existing information databases highlight some limitations of these automatic image-generation methods. The rapid construction of increasingly realistic digital images draws new boundaries between real and unreal, highlighting a strict relation between text and image regarding semantics and logic descriptions. Therefore, a critical look into the use of these applications as tools for a reliable representation of architecture becomes cogent. We started from representations established by the treatises, as in the case of architectural orders. Most of these drawings, proportions, and rules are derived from descriptive parts in Vitruvius’ text. However, the graphic interpretations result from the architect’s experience and culture, which has become the basic grammar of architecture. This research stems precisely from the connection between Vitruvius’ text, the new text-to-image contents, and the established representations of the treatise writers. The comparison considers both image reliability and geometric rules, testing the current potential of GPT systems for image creation and reliability.
Digital Echoes—Revisiting the Venetian Church of Crociferi from the Perspective of Artificial Intelligence
This work investigates the application of artificial intelligence (AI) technologies within the framework of the discipline of architectural representation. It challenges the typical AI workflow, which often applies AI to unknown territories resulting in unpredictable, yet too easily accepted, outcomes. The case study of the experiment, partly described here, is the lost Church of Santa Maria Assunta dei Crociferi, in Venice, Italy. Despite the lack of scholarly attention, this site is noted for its historical significance. The experimentation begins by analysing the research made by two architecture historians: Sherman (The Lost Venetian Church of Santa Maria Assunta dei Crociferi: Form, Decoration, Patron-age. Independent Publishing Network, London, [1]) and Galeazzo (Dinamiche di crescita di un margine urbano: l’insula dei Gesuiti a Venezia dalle soglie dell’età moderna alla fine della Repubblica, Università Ca’ Foscari, Venezia, pp. 455–692, [2]), then DALL-E is used to generate images based on the same textual information, adapted to prompts. Finally, the results are evaluated in relation to traditional approaches belonging to the methodologies of history of architecture and architectural drawing. The aim is to observe how the AI model responds to the gaps in visual documentation, without necessarily expecting a precise reconstruction of the church. A key factor impacting this methodology is AI’s real-time evolution, which may shift research parameters. Most of all, this study discusses what the AI-generated images reveal about the databases used to train these models, as the experiment identifies patterns that reflect aspects of architectural knowledge widely available online. The paper investigates the still blurred connections between AI research and architectural practice, approaching both possibilities and limitations, providing architects with critical knowledge and enhancing awareness of their newest instruments and tools.
The New A.I.: Gaining Control Over the Noise
The paper investigates the evolution of AI-assisted architectural representation from early unpredictable text-to-image generation systems toward more controlled and precise workflows based on Stable Diffusion, ControlNet, and LookX AI. The research analyzes the role of latent spaces, diffusion models, GANs, convolutional neural networks, and AI rendering systems in architectural visualization, emphasizing the transition from exploratory AI image production to controllable design-oriented generation. Through experimental workflows combining Rhinoceros 3D models, ControlNet preprocessors, segmentation maps, depth maps, edge detection, and prompt engineering, the study evaluates how AI systems can support architectural rendering, stylistic control, spatial coherence, and atmosphere generation. The paper compares open-source and cloud-based AI platforms, discussing the balance between creativity, predictability, customization, and architectural precision in contemporary AI-assisted design workflows.
AI Text-To-Image Procedure for the Visualization of Figurative and Literary Tòpoi
The paper proposes a workflow for translating literary and figurative textual descriptions into AI-generated images through text-to-image neural networks. The research combines linguistic analysis, lexical semantics, prompt engineering, and Stable Diffusion-based image generation to investigate the relationship between verbal and visual representation. Drawing on theories of visual culture, ekphrasis, and Aby Warburg’s Mnemosyne Atlas, the study develops a methodological framework for guiding neural networks through semantic keywords, syntactic structures, contextual references, and prompt modulation. The workflow is tested on literary and architectural texts from different historical periods, including utopian cities, nineteenth-century urban descriptions, and imaginary urban narratives. The research demonstrates how AI image generation can support the visualization of literary spatial imaginaries while also revealing the ambiguities, arbitrariness, and interpretative challenges inherent in translating text into visual form.
Floating Acrobats: Exploring Exaptation in Architecture Through Artificial Intelligence
The paper explores the intersection between generative artificial intelligence, virtual reality, performance art, and architectural experimentation through the practice-based installation Floating Acrobats. The research investigates the concept of exaptation as a design strategy, examining how AI-generated imagery, VR environments, sculptural design, and performative spatial experiences can be repurposed into new artistic and architectural functions. The methodology combines hand sketches, 3D modeling, AI-generated visual simulations, immersive VR environments, and interactive exhibition design to create a multisensory installation centered on resilience, adaptability, and collective creativity. The study reflects on the transformative role of generative AI in artistic production, virtual exhibition spaces, authorship, and morphospatial experimentation, proposing AI as a catalyst for new forms of architectural imagination and performative interaction.