Bringing Italo Calvino’s Invisible Cities to life with AI
Written by: Kee Moon Jang and Fábio Duarte
Reimagining Invisible Cities
If you’ve ever read Italo Calvino’s Invisible Cities, it is not your typical book about urban landscapes. Instead of detailed maps and street names, Invisible Cities is a series of poetic dialogues between the explorer Marco Polo and the Mongol emperor Kublai Khan, wherein Polo describes 55 cities that are reflections on memory, desire, death, and other intricacies of human experience. Calvino’s narrative transcends mere storytelling, inspiring contemporary discourse on architectural design and urbanism that challenge conventional perceptions of urban spaces.
For decades, these cities have only existed in the imagination of readers. Many have tried to bring them to life visually with creative illustrations sketched from their unique interpretations. More recently, generative artificial intelligence (AI) tools like Midjourney and DALL-E have been used to create visual representations of these imagined cities. These, however, are often based on fragmented descriptions, distilling a few keywords or themes into an image that reflects the artist’s own interpretation.
But what if we could see and walk through the Invisible Cities as Marco Polo explored during his travels? Rather than piecing together an image from the hints Calvino left behind, we took on Marco Polo’s role: feeding the entire chapter of each city into DALL-E 3, allowing it to portray a picture from Polo’s perspective, just as he vividly narrates each city’s essence to Kublai Khan. Further, we bring the imagined cities to life using Sora, and invite the Khan to the Invisible Cities.
From text to image: Invisible Cities seen through DALL-E 3
We took Calvino’s poetic descriptions and fed them into DALL-E 3 using OpenAI’s API to visualize each of the 55 cities. The prompt was in the format as below:
“In the book Invisible Cities written by Italo Calvino is a conversation between the Mongol emperor Kublai Khan, and Marco Polo. The majority of the book consists of brief prose poems describing 55 fictitious cities that are narrated by Polo, many of which can be read as commentary on culture, language, time, memory, death, or human experience generally. Lets imagine Marco Polo had the generative power of DallE and were able to provide visual aids for the 55 cities.
Below is the full chapter of the city ‘{CITY_NAME}’ in the original book as narrated by Polo.
{ The original chapter of CITY_NAME }
Given this description, please provide a vivid style high-resolution 3D digital piece in Photorealism style, the photorealistic city landscape of ‘{CITY_NAME}’, detailing its urban features and characters. Do not include any text within the illustration, and be very accurate with the number of objects when clearly mentioned.”
The result was a stunning set of AI-generated images that transformed abstract texts of these fictional cities into realistic visual art. For example, take Diomira, a city of “sixty silver domes” and “bronze statues of all the gods.” DALL-E turned it into a shimmering metropolis that blends both Byzantine grandeur and futuristic elegance in architectural elements. The “golden cock” sits on a tower looking over the urban plaza that is lit up by “multicolored lamps (…) of the food stalls.” Another example, Thekla, is depicted as an endless construction site, where we can only “see little of the city, beyond the plank fences, the sackcloth screens, the scaffoldings, the metal armatures, the wooden catwalks hanging from ropes or supported by sawhorses, the ladders, the trestles.” In addition, workers are here and there across the scene, embodying the city’s belief that construction must not cease.
These AI-generated images serve not only as visual companions to Calvino’s text but also as standalone mediums that offer viewers with photorealistic, often breathtaking, illustrations that materialize their imaginations of the Invisible Cities. The generated images of all 55 cities can be found here.
Bringing Invisible Cities to life with Sora
While the image of Octavia portrays successfully the “spider-web city,” the Sora-generated video enables us to fly between two steep cliffs of the mountain where “the city is over the void, bound to the two crests with ropes and chains and catwalks.” The delicate web-like structure is brought to life, as if we were just above the “rope ladders, hammocks, houses made like sacks, clothes hangers, terraces like gondolas (…)” that are hanging precariously in Octavia. Adelma is even more impressive, where the video not only captures the dock at dusk, but also creates the hidden scenes as it flies over the waterway.
Seeing these cities in motion adds a whole new layer to the imagination of Invisible Cities. We are now stepping into Calvino’s novel, and exploring the cities just as Marco Polo, transcending space and time. They are no more invisible; they are now visible, and also alive.
AI’s role in shaping urban imagination and the future of fictional cities
Our little AI experiment wasn’t just about having fun (though, let’s be honest, it was a lot of fun). It also raises some big questions about how AI is changing the way we think about cities and design. For architects and urban designers, AI tools like DALL-E and Sora offer new ways to translate natural language into visual and animated forms that were once confined to pure imagination. The generative power can now let us explore new urban aesthetics and prototypes beyond conventional practice; not just sketching it, but actually seeing it in action before a single brick is laid.
What about beyond design? Many of Invisible Cities revolve around memory, desire or name, themes that are central to the place identity of cities. The outputs from our experiment demonstrate the representation of such subjectivities is now capable with AI-powered storytelling. Moreover, this would be helpful to democratize the creative process, serving as an effective teaching tool for individuals without formal design training to visualize possible futures of cities shaped by climate change, new technologies, or different ways of living.
Of course, AI-generated cities come with their own challenges. What biases are baked into the algorithms? Do we risk creating cities that all look the same, instead of embracing the diversity that make real places so special? In 2023, we already wrote in Dezeen and warned about AI making predictions “based not only on patterns of image data but also patterns of social stigmatization about certain urban populations.” Reliance on these tools may inadvertently reinforce biases present in existing data and lead to homogenized or culturally insensitive designs.
In the end, Calvino’s Invisible Cities were never meant to be seen; rather, they were meant to be felt, imagined, and interpreted differently by every reader. But by bringing them to life through the generative power of AI, we’re adding another dimension to that experience. We not only pay homage to the literary genius, but also we’re opening up new ways to think about not just fictional places, but the real cities we live in today and the ones we’ll build tomorrow.
So, if Italo Calvino was around today, would he approve of our visible and alive versions of Invisible Cities? It is hard to say, but we’d like to think he’d be fascinated watching as the cities he once dreamed of take on a life of their own.
