GANs And Deepfakes Could Revolutionize The Fashion Industry
Over the last few years, the field of artificial intelligence (AI) has grown by leaps and bounds. Researchers are working on driverless cars, voice-controlled smart assistants and image recognition that can spot tumors in photos better than radiologists.
On the retail front, AI is already changing how customers shop online. Algorithms are already suggesting items you might like based off of previous searches or purchases. The Echo Look is Amazon’s “style assistant” that takes a photo of your outfit and makes fashion recommendations that are conveniently available for sale on Amazon. And AI will transform online commerce for retailers in an even more major way in the near future — realistic digital models may eventually replace humans.
What Are GANs?
The technology that could make this happen is called GANs, short for generative adversarial networks. GANs are two artificial neural networks that work together to create real-looking media. The networks are both trained on the same dataset of images, videos or sounds. Then, the first tries to create new examples that are good enough to trick the second network, which works to determine whether the new media it sees is real or not.
It is a system only dreamed up in 2014, and it is already showing a lot of promise in a range of industries, from medical fields to particle physics. Unfortunately, it is becoming better known for a Black Mirror-esque use case. GANs power deepfakes, or videos that use AI to paste an unwilling participant’s face onto a performer in a pornographic video or a potentially repuration-degrading clip. But just like a hammer can be a tool to build a house or hurt someone, GANs are what we make of them. And for fashion, it will make shopping more personal and cheaper for retailers, and it could eventually increase sales.
Digital Models Are Already On The Rise
AI like GANs means we aren’t limited by what humans can do anymore, or reliant on people at all if we don’t want to be. Brands will be able to rent supermodels that aren’t real, essentially having designer humans made by AI. Already, there is Miquela, an AI fashion blogger with 1.4 million followers on Instagram. Besides regular postings showing her in the latest styles, she even collaborated with Prada during Milan Fashion Week, “wearing” pieces from the new collection.
Miquela is only the first of what will likely be many totally artificial models, but with GANs, we could blend the real and virtual. Celebrities could agree to be a spokesperson for a fashion line and model clothes without having to do a photo shoot. Brands could show outfits on a variety of models with different skin tones, heights and weights. Users might even upload photos of themselves that are then used to create custom digital models to preview how an outfit would look on them before purchasing. It’s the ultimate personalization.
This isn’t diversity for diversity’s sake. A 2012 study showed just how much higher purchase intent when clothing is shown on models who are closer to the consumer’s size, age and ethnicity. When the model displayed was the shopper’s size, their intent to buy increased 200%. If the shopper was larger than a size 6, that increase went up to 300%. Black women were 1.5 times more likely to purchase a product when it was presented on a black model. When shoppers saw models that were close in age, their purchase intent went up by more than 175%. Companies using GANs could craft sites for every individual shopper.
A wider variety could be better shown when suggesting looks, too. Styling a model with a full outfit increases purchase intent, but if any of the pieces are sold out, it can hurt overall sales. A GANs-powered system would let a stylist dress a digital model with alternative pieces already picked out that get swapped in quickly if anything goes out-of-stock.
Relying on digital models, according to my company’s market research and customer interviews, could also cut up to 75% of the cost associated with shooting products on real people. Companies can spend between $100 and $1000 for each look for the web. Brands often reshoot certain items that don’t seem to be moving well, compounding expenses. Beyond cost, it can take up to 60 days to take a photo, edit it and get it online. A fad has come and gone in that time frame. During Fashion Week, customers can often tune in and stream a designer’s show in real-time but then have to wait six months for the clothes to hit the shopping floor.
With GANs, the items would be shot once and broadcast almost immediately. The time to market is the center of the story for the fashion business in 2018. Zara has built an empire doing fast fashion faster and better than just about anyone. That same speed will be available to more brands with digital models.
The Future Of The Industry
There has been a lot of buzz around virtual fitting rooms and creating avatars to customer’s measurements. However, until those technologies are perfected and adopted by consumers, using GANs with the ability to visualize and even personalize how products are shown is a big step toward increasing purchase intent and reducing returns. These are small steps, but they’re steps in the right direction nonetheless.
Producing enough models with GANs to display all the inventory of a brand is still a technical challenge. But the demand is out of control, so it’s only a matter of time. Tech is challenging what fashion and retail can be. AI will reinvent shopping, one digital avatar at a time.