Is the fashion industry ready for the age of AI?

Virgil Abloh, the former Louis Vuitton men’s art director and the head of the trendy brand Off-White, who passed away last year, is still a legend in the fashion industry. In many speeches and interviews, he has revealed that in order to adapt to the ultra-fast rhythm of the luxury industry and the almost year-round state, all his design work is basically carried out through two mobile phones and the instant messaging software Whatsapp.

Virgil has a team that helps him transform from concept to design. He collects fashion “materials” that he stumbles upon and recognizes, takes pictures with his mobile phone and sends them to his design team. The design team draws inspiration from the photos he sends, designs them in sketches or realistic renderings, and sends them back to him for revisions. Relying on the strength of the design team, Virgil Abloh can undertake the high-intensity creation of at least six series of the two brands every year.

Such “ghostwriting” practice is actually not uncommon in the fashion industry. Well-known fashion designer Miuccia Prada designs her ideas by “translating” her ideas; Raf Simons will prepare archives containing images and inspirations as the basis for the team to make sketches; Rei Kawakubo also said that her designs “mainly rely on Words and imagination are done.”

Looking back, designing with words or referring to images is not exactly where artificial intelligence (AI) comes in. DALL·E is a leader in the new generation of artificial intelligence models, and its essence is a “text-to-image” model. People can create very realistic and detailed images in a very short time by entering text. This is where the “AI painting” that exploded on the Internet at the beginning of this year came from.

On October 20, CALA, an integrated platform for designing and producing clothing, announced that it has added AI tools based on the DALL·E model to its platform, which is a first in the fashion industry. Similar to DALL·E’s rules for generating images, users can generate a variety of clothing design results in the system by simply describing in language or uploading reference images to the platform, making clothing design so convenient and simple. CALA co-founder and CEO Andrew Wyatt said: “Since most of us don’t have a world-class on-call design team, the hope is that CALA will be a platform that will make it easy for anyone to use human resources at any time.” Smart technology for design.

AI helps break creative resistance

How does CALA’s AI tool achieve design “ghostwriting”? On the web page, users can choose one of the dozens of product templates provided by default as a base for their designs. The category of products is not limited to clothing, but also includes accessories such as handbags. In the text field, the user first enters an adjective describing the desired appearance of the product, such as: “dark, delicate and velvet”. Then add your desired detail keywords to the trim and features section, such as: “sewn logo patch”.

After the text is set, click Generate, and the program will display the results of the six created products to the user for selection. If you are not satisfied with the generated results, the user can also use the “Regenerate” function to generate the design infinitely based on the initial keywords until a satisfactory result occurs, and the generated design can also continue to be edited and modified for a second time. Some brands will also re-upload existing designs into the system, and generate six new designs that are slightly different from the original design, as a source of inspiration for “series derivatives”.

Andrew Wyatt noted: “We often see creative bottlenecks for even the most experienced designers, or people who want to create a brand but lack the artistic or technical knowledge to bring their ideas to life, and our platform will eliminate those Barriers to creation. DALL·E API is very powerful and can change many rules of the game. Inside the company, we are even able to use it to help us build 3D product template rendering kits. We are very happy to see that people can use these powerful artificial Smart tools for design.”

Although the creative function is very powerful, the reporter found that the current AI tool function of CALA still has many limitations in creation. For example, the outline of the product template provided by default cannot be changed, and more basic instructions like “shorten the sleeves” still cannot be implemented. But I believe that with the continuous development of the DALL·E model, CALA’s AI tools will also have a lot of room for improvement.

Portugal-based fashion company Fashable has also developed its own artificial intelligence using Microsoft Azure machine learning and the open-source framework PyTorch. They trained it by collecting images from e-commerce sites and social media, allowing it to automatically generate new designs based on past fashion paradigms. Orlando Ribas Fernandes, co-founder and CEO of Fashable, said that with artificial intelligence-aided design, brands can more easily create seasonal fashion collections, more efficiently choose whether to produce certain items, and avoid wasting labor at the same time. It also greatly saves capital costs.

Mining the fashion market value of AI

The origins of AI can be traced back roughly 70 years, but corporate adoption of AI has actually only taken hold in recent years. In the past one or two years, it is not difficult to find that once artificial intelligence technology finds its own suitable position, its utilization rate has also surged visibly, and “full bloom” in our lives.

The combination of artificial intelligence and the fashion industry is not unfamiliar. In the past, artificial intelligence was mainly applied to the e-commerce platforms of fashion brands, responsible for intelligent customer service, or providing simple personalized product recommendations to shoppers. However, as consumers’ needs for personalized experience become more and more diverse, and the need for enterprise inventory management is also growing, artificial intelligence is now playing a more sophisticated role in the fashion industry.

In terms of shopping experience, last year, fashion retail platform Farfetch demonstrated an artificial intelligence technology platform called “Future Store”, which provides stores with a shared customer purchase history database, which can record consumers’ online to offline purchases. The information and data of all purchasing behaviors, through in-depth insight into customer needs, ultimately promote a super-personalized consumption experience; in terms of customer management, Mytheresa, a well-known German luxury e-commerce platform, uses artificial intelligence to predict fashion trends and cultivate customer loyalty; In terms of inventory sales, LVMH just announced a five-year strategic cooperation with Google last year to jointly develop artificial intelligence-enabled business solutions to improve business operations in areas such as demand forecasting and inventory optimization.

There are various signs that artificial intelligence has penetrated into the fashion market. Not only has it taken a place, but the scale of artificial intelligence in the fashion market is also growing rapidly. According to data, on a global scale, the market size of artificial intelligence in the fashion market is expected to grow from $228 million in 2019 to $1.26 billion in 2024, with a compound annual growth rate of 40.8% between 2019 and 2024. rising.

Rediscover the meaning of exploring beauty

However, in addition to enjoying the convenience and thoughtfulness brought by artificial intelligence, there are more questions that “land” along with it: Does the design generated by imitating the pictures and words in the database involve infringement and plagiarism? Will artificial intelligence replace human designers? How will the work of human designers change in the future?

In this regard, Wyatt said that the reason why they initially adopted artificial intelligence tools built on DALL·E is that OpenAI, the development company of DALL·E, has always adhered to the core principle of developing “responsible artificial intelligence”, and generated under this framework. ‘s design results do not replicate other branding’s iconic graphics.

And Orlando Ribas Fernandes of Fashable also said that their artificial intelligence has been programmed to never copy the iconic pattern of other brands, so even if a user enters another brand’s best-selling product as a keyword, the final result will be mixed. Incorporate original elements.

Referring to the challenges of artificial intelligence to human designers, Pedro Domingos, a professor of computer science and engineering at the University of Washington, pointed out: “Artificial intelligence can create music, write stories and create art, and fashion design is also natural. However, most likely to happen The situation is not that artificial intelligence will completely replace designers, but become an indispensable tool for them.” Just as world-renowned architects such as Zaha Hadid need computer-aided modeling design in their work, “Having artificial intelligence of fashion designers are also able to come up with passionate ideas. AI just amplifies their creativity, not replaces them.”

Technology is developing, and every progress will bring fashion to a new starting point. Consumers, designers, and brands need to constantly adapt to their new roles. But we understand that when the power to create beauty is “dissolved” from the hands of designers to everyone, modern fashion design is no longer a simple line from “creation to production” – it re-empowers everyone to explore beauty , to discover the meaning of beauty.