If you didn’t already know artificial intelligence was used to design Collina Strada’s Spring/Summer 2024 collection, it would be all but impossible to tell.
At last Friday’s runway show, set amid a rooftop farm on the Brooklyn waterfront, the clothes betrayed no obvious hints that founder Hillary Taymour and her team spent weeks feeding images of past Collina Strada looks into the AI image generator Midjourney, guiding it with text prompts of varying complexity and diligently refining the results to produce the collection. The outcome was quintessential Collina Strada: a frenetic clash of shapes and fabrics in acidified colours coalescing into 21st-century psychedelia. Not even the show notes mentioned it, focusing instead on femininity as a source of strength in the face of the climate crisis and the curtailing of reproductive and trans rights in the US.
“I just think it doesn’t fucking matter,” Taymour said with a laugh in an interview this week. “There are so many other elements and layers to my show that it didn’t need to be about that.”
The use of generative AI tools has gathered momentum in fashion over the past several months, even amid concerns the technology could displace creatives and infringe intellectual property. Revolve released an AI-generated billboard campaign and capsule collection, backed an AI Fashion Week and is experimenting with it to design its lucrative private brands. Casablanca used it in the marketing campaign for its Spring/Summer 2023 collection. Others are turning to it for jobs like writing product descriptions.
Collina Strada appears to be the first designer label to use it so heavily in a runway collection, or at least the first to acknowledge it. Taymour called it a “game changer.”
Multiple times during the conversation she described it as pushing her brain further creatively. Where she found the AI useful wasn’t in producing flawless, finished looks but in how it could reimagine familiar ideas in unexpected ways, yielding results she might not have considered otherwise. (It’s a viewpoint others have expressed, too.) She likened it to being a child and not being limited by the learned ideas that become entrenched about how things should appear.
That played out in the collection in several ways. It informed the shapes in the clothing, like a corset dress that dripped down in the front as if to mimic the silhouette of a bodysuit. AI was also behind the prints, including a glitchy plaid that seemed blurred with watercolour florals, while AI-generated stars with distorted limbs appeared on clothing and as jewellery. Even the runway sculptures by David Aliperti were developed with AI.
Taymour, who was introduced to the technology by her art director, Charlie Engman, emphasised that it was not a shortcut. She said she spent several weeks “using AI 24 hours a day, seven days a week, literally just prompting all the time.” She would start by uploading previous Collina Strada looks, sometimes blending two together. The text prompts she wrote telling the AI what to do could range in length from “a word to a novel,” she said. Each time the AI returned four generated images, and each time she would choose the best one, refining repeatedly until she arrived at something she liked. For one dress, she said she went through 210 versions.
“It’s not like you’re not doing the work,” she said. “It’s not like it’s the cheap way out.”
Turning the digital renderings into physical garments was its own challenge. The shapes and textures could be difficult to reproduce. Her team had to source fabrics to match and needed to decipher where to put seams and how to construct the pieces. To figure out how to make the sleeve on one dress perfectly took four weeks. It didn’t help that the AI images only showed a single angle, so while Taymour could generate the front of a dress, she and her team still had to design the back themselves to produce a finished three-dimensional piece.
The labour involved and human ingenuity required should allay fears that AI will put human fashion designers out of a job anytime soon, in Taymour’s view. The issue has understandably caused a good deal of anxiety. After all, when businesses are given the option to reduce costs, they generally will, even if that means cutting human workers.
Many creatives are worried their work is being used to train the AI models that threaten them. The image generators and language tools like ChatGPT proliferating today are powered by models trained on huge volumes of data pulled off the internet. Already a number of artists and writers have sued AI developers alleging the technology infringes on their intellectual property rights because it was trained on their work without consent. While US law typically judges IP claims based on the final product, not the process, AI brings the matter into new territory for the courts to navigate.
Taymour said she started from existing Collina Strada imagery to avoid this issue, and to make sure the collection read as hers, not a derivative of something else. But she suspects other companies will use AI to generate their versions of looks from other brands. Still, she’s not overly concerned about it, because she said the AI’s results vary widely from one user to another, even starting from the same image.
The technology took some time for her to learn. As she succinctly put it, “it sucks for a week.” The real test is whether she now considers it part of her toolkit as a designer. Would she use AI in the future?
“It’s like asking me if I want to use a pencil,” she said. “It’s only going to keep getting better.”