Brands are unleashing generative AI design tools for customers
By Maghan McDowell, Vogue Business
Starting today, anyone can create their own digital Reebok sneakers using generative AI. It’s part of a surge in brands using the tech to put design power in the hands of their customers.
Through a new Instagram account, @ReebokImpact, users are invited to submit images to a chatbot to see them turned into patterns on sneakers, selecting from Reebok Pump, Classic or Club C models. They can then tweak the design, with the original image printed on the sole. People can then elect to buy the digital sneakers for $8 to wear in multiple games (including Fortnite and Roblox) and receive the 3D files, which work with Unreal Engine software and can be digitally tailored on one’s own image or on digital avatars. Anyone who buys the digital version also gets a discount code (up to 50 per cent off) for a standard physical version of that same style.
The emphasis is on generating something that wouldn’t physically be possible — such as wild, on-demand one-of-one patterns and colours, and full images under a glass sole — thanks to generative AI. The project is a partnership with metaverse and Web3 studio Futureverse. “Real life obviously has some sort of ceiling to the mechanics of innovation, but the metaverse can allow for this infinite imaginative opportunity,” says Futureverse co-founder Shara Senderoff.
It “marks an important milestone in our commitment to innovation and technology”, says Reebok CEO Todd Krinsky, adding that it is a way to “explore new and engaging ways to bring our brand’s ethos to a wider audience”.
Also this week, lingerie brand Adore Me will be expanding a pilot that enables customers to design their own bra and panty set. Called AM By You, customers can enter a text prompt to generate a pattern in one of three artistic styles that is displayed on an underwear set; once they are happy with their design, they can buy the physical version for $55. “This is the true promise of generative AI coming to life, because it’s one thing to automate a product description, but it’s another thing to make people feel like, without any previous talent as a fashion designer, they can actually imagine and get exactly what they want,” says Ranjan Roy, SVP of strategy at Adore Me.
In recent months, Nike has worked with 13 athletes to co-create AI-generated sneaker designs. The project, called A.I.R., turned generated images into physical footwear designs through hand sketching, 3D sketching and computational design, then 3D-printed the final products, with the goal of introducing this concept to consumers “in the near future”, Nike president of consumer, product and brand Heidi O’Neill told Vogue Business during a recent presentation in Paris. Versace, Loewe, Gucci and Balmain have also experimented with crowdsourced generative AI artworks or products.
Generative AI has enjoyed a generous hype cycle in fashion and marketing because of its ability to enable most people to quickly create professional-looking, artistic images and videos with simple prompts, often inviting a playful call-and-response process that provokes engagement and delight. It can also produce fantastical content that wouldn’t normally be accessible. For brands, this translates into a practical use of generative AI, which can personalise products at scale and can be implemented relatively quickly. By putting the reins in the hands of consumers, brands are also able to experiment with less scrutiny — while collecting data on what sticks.
“Sometimes with emerging technology, we talk a lot about five years, 10 years, and we ignore the practical applications,” says Becky Owen, CMO of Billion Dollar Boy, who heads the agency’s innovation unit, Muse, leader of the AI experiments across Versace and Loewe that invited creators to generate social media content featuring their products.
It’s also a way to capitalise on the zeitgeist of AI curiosity while reinforcing a brand’s popular styles or marketing new products. “Our view was, we want people to experience this,” Senderoff says. “A lot of people don’t know what AI is. They have theories — ‘Does AI hurt my creativity? Does it help my creativity? How do I even access it?’ — but few have the accessibility to play around with it. We’re massive believers in the collaboration theory of AI versus the displacement theory that a lot of people have.”
DIY designs drive engagement and personalisation
Reebok’s experience lets each user generate four free digital shoes before making a purchase, and if they buy one, they can start the process again. The tool is accessible via a direct message on Instagram, usually a specifically designed AI chatbot, to hopefully enable a process that is natural and accessible, Senderoff says. People who pay to download the digital files also receive them in a crypto wallet as an NFT, but this element is intentionally kept under the hood, she says, primarily to prepare for an anticipated interoperable future, in which people can own and wear the same digital fashion item in multiple virtual spaces.
Reebok is hosting a contest to physically produce at least 30 one-of-one physical shoes inspired by top digital designs, but a future challenge will be considering how to scale phygital production.
Creator agency Billion Dollar Boy has found content that intentionally looks ‘out there’ as a key to success in the “engagement rate wars”, as Owen puts it. This is why she advises against using AI to recreate humans in images, but rather for creating something that is recognisably outlandish. Versace saw an increased video play rate of 1,460 per cent on the AI-generated social media videos, compared to its standard content. A recent survey commissioned by Billion Dollar Boy, of US and UK-based creators, found that 81 per cent of creators reported more favourable consumer engagement on their generative AI content compared to their traditional content.
There are other concerns, from both brands and consumers, about transparencies and disclosures that circumvent accusations of being misled, or taking jobs from someone who would normally be creating the content. Already, tech platforms including TikTok, YouTube and Meta have created disclosure requirements, and Billion Dollar Boy advises all brand clients to be transparent about which content is made using AI. Billion Dollar Boy also limits what intellectual property can be fed into AI models, and often, a brand’s products are manually added to the content after the original AI-layer is generated, partly to protect the brand’s IP, as rights and licensing issues remain somewhat unclear, she adds.
Adore Me has taken a fairly conservative approach to content moderation, SVP Roy says, with all designs fed through an automated review first — especially looking for prompts that include brand names. To build the tool, Adore Me custom-trained a Stable Diffusion model using AI-art generator Leonardo AI, which helps translates prompts not onto squares of art (as seen with Dall-E, for example) but rather patterns that are inherently compatible with the shape and context of a bra and panties. It prints the designs using a ‘Promptly Printer’ from Sri Lankan manufacturer MAS within one of its fulfilment centres, and the product is shipped in the same time frame as standard orders.
So far, early trends are shamrocks and Saint Patrick’s Day-related designs; skulls and witchcraft; and border collies and Australian shepherds. Since launching the small pilot in January, 1,000 customers have received products, with the average person generating 5.3 prompts (customers have unlimited prompts and aren’t obligated to buy generated designs). Adore Me also found that for most customers, this was their first experience using image-generating AI. “From a data perspective, at some point that could inform planning and production,” Roy says.
There are already plans to scale, with different styles of blank products available to be printed on, and the in-house team is already debating whether it should financially categorise the cost of prompts as cost of goods sold, or as a marketing expense (prompts require considerable computing power and can become expensive).
The Reebok pilot is just the beginning, and it signals a long-term partnership with Futureverse. The NFT element, the digital wearables, and the conversion to physical are key details on the roadmap. “What happens in this evolution of the internet, where the first iteration is maybe you get a discount and special access, but how do we go from a brand selling digital versions of their stuff to consumers being able to do what the data shows us they love?,” Senderoff says. “We know that consumers want to leverage their digital fashion sense to inspire their physical, and we know that the physical wants to inspire the digital.”