In yesterday's post, we explored Francesco Guardi's works, focusing on his capricci, imaginative compositions where reality and fantasy converge to form idealized architectural visions. These paintings and drawings combined together existing buildings, evocative ruins, and fictional structures, bathed in dramatic light to create idealized scenes.
This technique of blending disparate elements into a new landscape echoes the modus operandi of Artificial Intelligence.
Consider how text-to-image models generate fantastical architectures by recombining fragments from various sources they have been trained on, melding influences from different styles, periods, and locations into an entirely new composition. Like Guardi's capricci, these AI-generated creations remix reality, assembling structures that never truly existed but feel compellingly plausible.
In digital and generative art, algorithms act therefore as modern-day capriccio painters, iteratively reconfiguring architectural fragments into endless variations of surreal cityscapes.
A user's input prompts the AI to orchestrate a fresh vision, an ancient Roman ruin reimagined within a cyberpunk metropolis, a Gothic cathedral adrift in the sky, or, much like Guardi's Venice, a city on the water that exists nowhere but in the image itself.
Experiments with Midjourney, using textual prompts inspired by Guardi's Capriccio with Ruins and Classical Ruins (View this photo) and Architectural Capriccio: A Palace Colonnade (View this photo), without directly feeding the images into the AI, produce grand yet subtly uncanny compositions.
The results suggest a Venice of dream and distortion: conglomerates of houses adrift on the water, domes improbably perched atop bridges and slender columns too delicate to bear the weight of the buildings they support. The light in the images produced by Artificial Intelligence is eerie, evoking Guardi's Rialto Bridge after the Design by Palladio, but with a gossamer, otherworldly quality.
Yet it is precisely in experimenting with Artificial Intelligence, prompting it to generate capricci, pushing it toward new iterations, and coaxing it to refine its visions, that one realizes how closely this process mirrors that of the artists themselves. Many historical capricci were, at their core, acts of iteration, painters returning to the same motifs, rearranging familiar architectural elements in ever-new compositions - a ruined archway might appear in different landscapes, a colonnade reconfigured with subtle variations, each piece both a repetition and a reinvention.
Thus, capriccio and iteration converge in their shared ability to recombine the known into the unexpected. Iteration provides structure, a framework of continuity, while capriccio injects playfulness, allowing for creative distortions that make each variation feel fresh, even surreal.
In the previous post, we highlighted how the capriccio approach is already embedded in contemporary fashion (think about Renaissance sleeves reinterpreted in a modern silhouette or 18th century corsets reimagined through a punk aesthetic...). But iteration, too, plays a defining role.
Sometimes, reinvention does not require borrowing from past eras; instead, it emerges from within, by revisiting an archive and reshuffling its codes. Fashion thrives on this process of reiteration and recontextualization (consider Alessandro Michele remixing Valentino's legacy designs for his sophomore Haute Couture collection at the Italian house). Iteration turns therefore not into mere repetition but into a means of weaving past and present into something entirely new.
For example, for their latest Haute Couture Spring/Summer 2025 collection, Viktor & Rolf transformed their own archive into a dataset, generating a myriad of variations on a single outfit - a beige trench, a white shirt, and blue trousers - all crafted in silk gazar from Ruffo Coli, a storied family-run textile house in the Como silk district. This fabric served as the trait d'union, a common thread weaving through the collection's many iterations.
During the show that took place at the end of January in Paris, an AI-like robotic voice repeated: "Beige trench in silk gazar, white shirt in silk gazar, blue trousers in silk gazar."
Each time the results were different, though at the same time containing elements from previous collections: the main fabric employed referenced Viktor & Rolf's Haute Couture Spring/Summer 1999 Blacklight collection, originally made from the final rolls of silk gazar developed by Cristóbal Balenciaga in the 1960s with the now-defunct Swiss textile producer Abraham.
Yet beyond this direct material reference, the silhouettes recalled a rich lineage of past collections: the inflatable silhouettes of their Autumn/Winter 1998 collection; the sculptural collars of the Autumn/Winter 2003 designs, the exaggerated bows of their Spring/Summer 2005 collection, and the dolls of their 2008 Barbican exhibition.
Each new look was an iteration: shirts transformed into voluminous capes, trousers widened into sweeping palazzo cuts or shrunk into slim silhouettes, trenches morphed into opera-length coats covered in bows or turned into tiered gowns.
Through this exercise in repetition and reinvention, while the description (or prompt…) remained the same, just like the fabric, the results, based on the house codes, were always different, like shifting capricci, altered by mathematical variables - volumes, proportions, textile manipulations, presenting fashion as an iterative process, where creativity emerges from the tension between repetition and transformation, a juxtaposition that Viktor & Rolf have been exploring extensively in recent collections, drawing heavily from their own past designs. Each new iteration becomes a deliberate exercise in self-referencing, a controlled capriccio where past and present collide.
In the end, the question remains: is the human approach to capriccio and iteration superior, or does AI offer a more compelling alternative? There is no definitive answer, it is, ultimately, a matter of personal perspectives. But if you're a fashion design student, the best way to find out is to experiment. Push the boundaries of iteration, embrace the capriccio, and see where these processes take you.
Comments