As seen in yesterday's post, we are experiencing a boom in the usage of Artificial Intelligence (AI) in different sectors, including healthcare. Fashion-wise we have seen some attempts at using AI text-to-image apps on the runway during the latest fashion shows, namely in prints (Christopher Kane) or images re-elaborated with a human touch (Coperni), but also in advertising campaigns (Moncler) and look books.
During Paris Men's Fashion Week, for example, Vuarnet used AI for a look book in which an IRL image was juxtaposed to an AI generated interpretation of the same image in a "natural" environment.
As the frenzy around NFTs is dispelling, a new star, Artificial Intelligence, is on the rise and the technology is also progressing extremely fast with image generators such as Midjourney and Stable Diffusion often prominently featured in the news on a daily basis.
One of the most surprising things is that some artists producing fantasy images of unlikely collaborations and imaginary collections are coming up with fictitious products that look more desirable than the original ones: when Nike dropped the first images of its Tiffany collaboration, many fans of the two brands were disappointed at the banality of the final product - a black Nike sneaker with a Tiffany green swoosh. It didn't look unique nor desirable.
In the meantime, Midjourney AI artists such as Rickdick produced fantasy sneakers in Tiffany green covered in sparkling gems or in intricate three-dimensional flowers, looking not just innovative and fun, but extremely desirable. As the Midjourney or Stable Diffusion aesthetics spread, impacting on consumers and changing tastes, the debate about AI intensifies.
The main fear is that AI tools will replace human jobs: but, as seen in the previous post about healthcare, Artificial Intelligence tools produce the best results when used alongside doctors.
In the same way, in the design field, these systems are better employed when they are at the service of designers who may benefit from the help of other professional figures such as prompt engineers (by the way, it would actually be refreshing to see more young women getting into this sector and becoming AI specialists).
AI text-to-image platforms seem very straightforward to us, we think we just have to write down a simple prompt and some extraordinary image or garment in the case of fashion prompts will appear on our screen almost by magic, but that's not how things work.
A well-written prompt is of vital importance if you want to come up with great results and very polished images. An example of how intricate a prompt can be was offered by Reddit user CurryPuff 99 who recreated a photographic interpretation of the famous "Lofi girl" Japanese-style animation. A step-to-step guide of this process was posted on the Eluna.ai Instagram account.
The results are stunning and prove the prompt is not just a questions of typing down a mere description such as "Studying girl with stack of books and brown flower pot on table, brown cat on white window ledge", but of being able to add notes on the CFG scale, the size of the image, mask blur, denoising strength and so on.
If design teams at fashion houses are interested in employing AI images in their work and pursuing such levels of precision, the most powerful fashion houses will have to establish their own AI department (mind you, some of them already have it…) where they can train a neural network with their own archival images (an archivist and fashion historian would be handy, obviously).
Another asset for a fashion house could be a prompt engineer who could lead the AI and allow it at the same time to express itself generating a new aesthetic for the fashion house in question.
This relatively new figure will have to work together with the designer, but also with the in-house graphic artists: images generated with AI software but without using original archival images of the house in question, may prove borrowed or "inspired" to collections of other houses and will have to be scrapped; images resulting from blends of archival pictures may instead need some graphic interventions, but may in the end generate something new.
The "/blend" prompt in Midjourney allows, for example, to mix together two or more images and this fairly simple blending exercise may be the first step towards creating something new or an experimental look. In this post there is an example of a blend of Dior's iconic bar suit with a design from Raf Simons' Dior A/W 15 collection.
The first raw results may not be convincing, but they can be tweaked, upscaled, remastered, altered and in the end Photoshopped by a graphic designer.
Such exercises - when supported by a neural network trained with a robust selection of archival images - may even lead to capsule collections of blended looks or to blended pictures that may be used in an exhibition as well.
But there are other changes that may be implemented in fashion companies: some may indeed opt to speed up their own systems with other deep learning models. Some companies recently turned to Cala's new artificial intelligence tools.
Built using OpenAI's advanced AI system DALL-E, fashion supply chain platform Cala allows to generate a variety of designs and alter them with text descriptions or upload personal images and come up with an impressive variety of apparel and accessories.
The system has a free option or a subscription plan: you start by selecting a template or a product (the range is wide and includes outerwear, dresses, sweats, hats, swimwear, bags, socks and so on).
The next step consists in adding the language prompts in the "Adjectives and Materials" section to describe colours, styles and materials (the images of the puffer jacket in this post were generated by the prompt "rainbow coloured, geometrical, hexagons, allover, repeated pattern"), adding more specifications in the "Trims and Features" section ("pockets", "hood", "zipper", "logo patch" for example). The next step is to choose if the system should generate a photorealistic image or a flat sketch.Cala will generate six examples of products, but, if you're not satisfied with them, you can click on "Regenerate" and you will be able to go through hundreds of versions and iterations to try and tune the designs obtained with what you had in mind. As an alternative, you can also upload an image of a garment you've created and DALL-E will return six images with slightly different variations of the original design.
An entire design team can work together on the Cala platform and, once you're finished, you'll be able to start producing the item you created as well. Cala offers indeed production facilities through its ghost team of designers who create patterns. The company is also linked to 20 factories located in 13 different countries.
In a nutshell, a system like Cala can facilitate the design process from product ideation through order fulfillment (please note that prices vary according to the items you order), all in one go and in a few minutes. The platform works with all sorts of brands: from large established fashion houses to independent designers. Besides, Cala also offers a mobile app (for iPhone) that allows users to design on the go, following the creation process from start to finish on a mobile device.
Cala promises users they will be able to run their fashion houses and, while we do know that running a fashion house is no mean feat, the platform seems to lift a lot of stress off the shoulders of designers with its fully automated process. There's secrecy on the designers and celebrities already using it but, according to the platform, there are already quite a few ones.
In a way these processes are all part of a digital transformation of the fashion industry: fifteen years ago, we were amazed by the possibilities of digital printing on fabrics; then came 3D printing technology that went from the production of rigidly unwearable clothes to soft and perfectly wearable pieces. Covid-19 pushed the digital revolution towards new worlds populated by avatars and digital wardrobes, but, in this post-pandemic state, we are pursuing new developments through AI.
But if the fashion industry is mutating, so is the role of the creative director who, rather than designing, is focusing on editing products, curating collections and maybe spinning a cohesive story around them (you may argue this was already the function of the late Virgil Abloh and it is the role that Louis Vuitton has in mind for his current successor at the French house, producer, songwriter, philanthropist, and entrepreneur Pharrell Williams).
Among the other entities gravitating around fashion houses and going through some changes because of AI tools there are advertising agencies. Using Artificial Intelligence is helping them to generate quick ideas and reducing the time dedicated to drafts. Educational institutions as well will have to go through some changes: balancing manual and technological skills will be a key point for students as some jobs at entry levels in fashion houses will be eliminated by the advent of AI departments and those who already know how to navigate this reality will be facilitated in finding other positions. Learning to balance manual and technological skills will be an asset: no AI is able to carry out manual processes, but AI tools also need a human to provide us with the best results, so improving both skills will be of paramount importance.
It is obvious that all these processes will imply a change in copyright laws as well: we have seen how the US Office didn't grant the registration of a comic created with the support of Midjourney, yet prompts are original concepts, they are the results of a set of skills provided by a human giving the machine an order, and it takes time to alter and hone the final picture, a process made by the machine, but triggered by a human being.
Who will be completely dropped by fashion houses as AI applications become more popular? Probably influencers who don't have any side skills apart from their looks represent the weakest link in the fashion chain of evolution at the moment. They can take all the Ozempic they want to fit in the sample clothes they are given for free, but fashion houses now have to feed several K-pop icons turned global ambassadors and sustain AI systems, platforms and teams to transition into a fast-paced digital future. For such houses it will be crucial optimizing their earnings and this will imply making drastic cuts.
Nowadays it is already possible to hire AI generated models (from Deep Agency that also offers the possibility of creating a virtual twin with an avatar that looks like you), and that's probably what some houses will do to save money. Where will the industry look for their out of the box thinkers who may turn into unlikely assets? Probably among the vibrant communities on the best Discord servers tagged with technology. Some of their members may not know much about fashion or look as hip as the characters populating the front row at fashion show, but, soon, knowing your technology will be trendier than peacocking outside fashion show venues.
Comments