Photo illustration by Chris Nicholls using Midjourney
Canadian fashion photographer Chris Nicholls recently conducted an experiment on Instagram where he posted 10 images of a young woman wearing a series of Baroque costumes and sporting some very strange headwear.
In one shot, the model – a Scarlett Johansson doppelganger – had a red crustacean sitting on her head. In another, lobster claws protruded from a white skull cap decorated with Chantilly lace, delicate white flowers and feathers. The images were highly stylized, surreal and thought-provoking.
But they were not created with a camera, a model, a stylist, or a hair and makeup person. Rather, they were machine-made: Nicholls was experimenting with a new artificial-intelligence platform called Midjourney, which uses text descriptions to construct AI-generated images that almost look like real photographs.
Similar to ChatGPT, which is capable of generating elaborate written responses on the basis of a few words, Nicholls used prompts such as “dramatic lighting,” “baroque dress,” “white flowers,” “lobster carapace” and “beautiful woman.” In 15 minutes, he had 30 visuals that the AI software produced by searching through its vast database.
After playing around with Midjourney – which Nicholls also used to create the visual of the woman on the Pursuits section cover of The Globe and Mail (see below right) – he concluded that the possibilities of the technology are both amazing and threatening. On one hand, he is awed by what this new era of “generative AI” can produce in the blink of an eye. On the other, he is outraged that it creates art standing on the shoulders of unnamed artists and photographers whose work it downloads for free.
“I’m conflicted,” says Nicholls, who for the past four decades has photographed some of the most famous people in the world including Kate Moss, Claudia Schiffer, Olivia Wilde and Taylor Swift. “The software is still fairly clunky and, at times, behaves like a toddler, jumping all over the map like a kid in a candy store.”
In one of his image experiments, the model was missing an ear. However, he believes it will work out such kinks, most likely at lightning speed. And when it does, he and many others in the fashion business worry that creative jobs will be dramatically changed, or worse, displaced.
“For the last five to 10 years, fashion companies have been struggling to keep up with the massive demand for content being asked of them on social media and other platforms,” Nicholls says. “As the technology gets faster and smarter, brands will see generative AI as a tool with the potential to create vast amounts of content at a fraction of the cost.”
And while some in the industry have expressed concern about potential job losses because of AI, Nicholls thinks that sentiment is naive.
“That train has already left the station. If you’re 22 and you want to be in the fashion industry, my advice is get on this as fast as possible because it’s what fashion companies are going to want.”
One New York-based model and futurist agrees.
“Up until this point, we kind of incorrectly assumed that creative roles and tasks are uniquely human. That made sense for most of human history, however that is no longer the case,” says Sinead Bovell, a Canadian who researches emerging technologies and educates people on their capabilities through her startup, WAYE.
“AI will eventually take my job as a model. Already, digital models and influencers are successfully breaking into the fashion industry from every angle, including in e-commerce, which supports a whole microeconomy of photographers, stylists, hair and makeup.”
While it is still early days for art and image generators such as Midjourney and its competitors, DALL-E and Stable Diffusion, Bovell sees a day in the near future when digital photo shoots will be common, producing covers for top fashion magazines as well as glossy branded advertising across all the social-media platforms.
The fashion industry is already moving beyond physical reality. During Paris Men’s Fashion Week in January, French sunglass brand Vuarnet relied on generative AI to create a look book. Retailers such as Replicant and DressX sell digital-only clothes. There are digital-only model agencies such as Diigitals. The website Generated Photos lets users create their own models with human faces. Fashion shows have been staged in the Metaverse. Even Marilyn Monroe has been revived as a virtual model to showcase the latest digital fashion from Balenciaga and Miu Miu.
As AI continues to evolve, more applications will emerge, predicts Kosta Koukoravas, founder and chief executive officer of Intelistyle based in London, England. His company uses AI-powered styling to personalize the customer experience of fashion retailers by styling their customers with the right clothes and outfits, online and in store.
“AI learns from all the examples that are out there. It can generate something new that has not been seen before based on what’s happened in the past,” he says.
Brands such as Zara, H&M, Dior, Macy’s and Nike all use AI in their business models, says Koukoravas, who adds advances in the technology will soon make it possible for brands to create realistic images of virtual garments and accessories based on customer demands and fashion trends. Brands will then share these high-quality visuals with consumers through social media or e-commerce platforms to get feedback before sending clothing designs to manufacturers – potentially cutting down on the 13 million tonnes of textile waste that ends up in landfills each year.
For example, last October, Cala, an all-in-one platform for designing and producing clothes, unveiled a new tool powered by DALL-E that can generate new clothing designs from text descriptions.
What does all this mean for the future of fashion? Koukoravas sums it up in two words: big opportunities.
“Creative teams will be able to use AI to come up with new designs, even getting end users and consumers involved in the process. You can imagine anything you like and ask AI to build it with you. Also, it will help fashion and e-commerce become more personalized. Suddenly, you could be looking at an image of yourself in that dress.
“It’s a big opportunity to make fashion more inclusive, more relevant to you. You can not only see models wearing these clothes, but you could see someone who looks like you wearing these clothes.”
By 2030, Price Waterhouse Cooper forecasts AI will add almost US$16-trillion of value to the global economy annually. For fashion brands, leveraging the ability of AI to exponentially increase growth will soon to be vital operational task, says Henry Navarro, associate professor with Toronto Metropolitan University’s School of Fashion. Even if AI-generated fashion images and virtual photo shoots don’t go mainstream in 2023, Navarro believes more fashion businesses will experiment with it, eventually paving the wave for adoption down the road.
“No industry has the luxury of staying the same, or even evolving slowly these days,” he says. “I empathize with those people working in fashion who are fearful of losing their jobs to AI. But those fears have been expressed before. They are the same fears painters had when photography came along. That film photographers had when digital came about, and that professional photographers had when everyone started carrying an iPhone.
“In every instance, their profession didn’t vanish, however, they did have to adapt to a new world with new possibilities.”
Consensus among fashion watchers is that major industry upheaval is likely five to 10 years away. However, the next generation AI text has already opened up a Pandora’s box of ethical, economic and legal questions. The lawsuits have already started, with stock photo provider Getty Images being the latest to sue Stability AI Inc. for misusing more than 12 million of its photos to train its Stable Diffusion image-generation system.
For the time being, Nicholls feels confident his job is safe primarily because generative AI still has a lot to learn.
“At this stage, it’s like fishing for beautiful imagery in a murky pond,” he says. “You set the bait and you keep pulling up images until it hits upon something beautiful – but it might not have hair or a nose, or it might have four ears.
“I spent two days sitting in front of a computer making the pictures I created for this newspaper. The process was fascinating, but I wouldn’t describe it as fun.”
Nicholls also recognizes that AI may have adverse effects on his industry, but he wants “to believe at the end of the day a human will still want to hire another human for the experience of working with that person.
“I became a photographer to travel and to have life experiences with other people. I hope to have many more.”
Generative AI systems have been grabbing attention with their ability to make images, text, music and more from a text prompt. We put some Canadian terms into three image AIs to see what they came up with, with some bizarre and surprising results.
The Globe and Mail