The model in front of me is wearing a white T-shirt and denim dress, showing a standard professional smile and posture.
If it is a shopping website, it is just one of many product photos. Consumers just scan the page mechanically, but if they take a second look, they may feel that something is wrong. The model’s limbs are slightly stiff, and the shadow under the neck is a bit strange. The boots are like plastic shoes worn by Barbie dolls. Yes, this is not a perfect AI model, but it will become more and more real in the future.
I am a model and I know that AI will eventually replace me.
Canadian model Sinead Bovell wrote this article for Vogue three years ago. Now, her prophecy appears to have come true.
AI models are invading
The AI model comes from the denim brand Levi’s. In March, it reached a cooperation with the AI model company Lalaland. This year, AI will be used to generate models. Regardless of authenticity for the time being, the birth rate and diversity of AI models are far ahead of humans. According to Lalaland, AI models are inspired by real people, but generated by AI, and their characteristics can be customized, including hairstyle, body shape, skin color, posture, emotion, etc. The process takes less than 5 minutes.
And many AI modeling companies have sprung up like mushrooms after rain. The start-up company Botika has launched an AI model service, which charges $15 per month, mainly for small companies. It has not yet started, and the waiting list exceeds 1,000. The official website shows photos covering the face, which can be applied to the heads of male models with different characteristics. It seems that the operation is very convenient:
Gone are the days when one model fits all. Just tap a button to change your model’s appearance based on expression, skin tone, hairstyle, background, and more.
ZMO, an AI-generated content website, shows that the folds and texture of the clothes in the diagram are well done, but there are no shadows to make the background look fake. At present, only the waiting list is open, but the promotional words make people want to queue up overnight:
Put product on models in minutes, not weeks!
The new Dutch startup Deep Agency calls itself an “AI photo studio and modeling agency” in a more high-profile manner, but the service is “cheap and good”: pay up to $29 a month, enter prompt words, adjust background, lens, pose, etc., and use AI It produces models, but it is stiff and plastic, and the price is more expensive than ChatGPT Plus.
Many e-commerce companies in China have also used AI tools such as Stable Diffusion to explore, and even re-developed AI model dressing systems based on these tools. In March, someone made a highly completed AI model map. The popular search topic on the Chinese social platform Xiaohongshu, “Is this the first job to be subverted by AI?” refers to Taobao models. But in the view of Simon Chambers, a senior model agency, AI models are still very limited, with neither emotion nor creativity:
AI models are more suitable for basic images such as references, rather than marketing or promotional aspects to build relationships with customers.
At present, most AI models can only take the basic route of middle and low level, but the important thing is the future possibility, no matter how rough it is, it has begun to be commercialized.
Original artist and Hollywood screenwriter’s successor
The reason for the continuous rise of AI models and start-up companies is simple. Where there is demand, there will be supply. After all, human models are not a small expense for SMEs. Tracy Porter is a self-employed clothing brand. In order to save money, she worked as a model herself. She also hired short-term photographers, models, hair stylists and makeup artists. However, it costs US$5,500 to take a product promotion photo. “Langzhu Huitou”, an e-commerce company that started AI models earlier in China, also came out and said: “This year we no longer need to spend 40,000 or 50,000 yuan a day to hire photographers and models… AI is completely enough .”
AI is aggressively invading the e-commerce business, which happens to be the most reliable source of income for models. Not to mention complete replacement, the emergence of AI has allowed the market to have more choices, which may reduce the treatment of human beings. Hollywood screenwriters and original artists are a lesson from the past.
Hollywood screenwriters are protesting against AI’s participation in creation. They are not afraid of being replaced by AI, but that the editors themselves write better, but are forced to rewrite the poor-quality scripts produced by AI, and their salaries will depreciate in disguise. On the other hand, many original painters have also become AI repainters. Party A feels that AI has already been used to produce pictures. The original painter only needs to make a simple repair, and the price can be kept very low. As a result, the original painter can only get an hourly salary for retouching.
(Source: Image by Freepik)
Some people now think that instead of using immature AI models, it is better to directly hire rookie models who have just entered the industry, and the average time and cost are not necessarily higher. AI models not only affect human models, the shadow of unemployment also covers the entire industry chain, including makeup artists, hair stylists, photographers and even planners. The spring issues of fashion magazines Vogue Brazil and Singapore have used tools such as MidJourney and Dall‧E to put AI models on the cover.
AI is proving itself, not only in e-commerce photography, but also in difficult fields such as fashion magazines. The first to bear the brunt are the models who can be replaced by AI models and are not yet unknown. Models in the United States and other countries are independent contractors. When the wave of AI models hits, it is difficult to protect them. This is what Os Keyes, an AI ethicist at the University of Washington, said. Ordinary models, photographers and artists are more likely to be affected by generative AI. Structuring Strength:
Apps such as Deep Agency, if effective, will further dislodge already precarious creative workers and shift money to such startups, where profits are more concentrated.
On the contrary, supermodels who dress up to attend the Met Gala and are active in the red carpet and big shows can temporarily feel at ease. In the face of technology, personal image still has value. CGI mannequins were an example a few years ago. They wear Prada, Dior, and Gucci, are personalized by human-made stories, and generate huge profits.
The established models are also planning for a rainy day. At the end of April, 50-year-old supermodel Eva Herzigová participated in body scanning, motion capture and other shootings, and wanted to create a digital clone “MetaHuman” for use in advertising campaigns. “We can completely change its appearance, from hair, clothes, makeup, poses to virtual environments, with the click of a button.”
But although it is very convenient and cheap for the owners, whether fans accept it is another matter. People may no longer believe that “there are pictures and the truth”, just like facing “arrested Trump” and “the pope in a fashionable down jacket”. A fan in the MetaHuman comment area left a message:
Miss your soul and the twinkle in your eyes.
AI models are not synonymous with diversity
Now there is no shortage of voices to support AI models, because it is a shortcut to show more faces, skin colors, and body shapes. Part of the reason why Lalaland CEO Michael Musandu started his business is that it is difficult to find suitable models:
As a Zimbabwean, I rarely see people who look like me in photos, commercials or shows.
But the diversity of AI models has also been questioned by people in the fashion industry. Levi’s, which announced the use of AI models, was burned because of this incident. People are angry that, with sales of 6.2 billion US dollars in 2022, they are always talking about diversity, but they are unwilling to hire models of all races. Levi’s, who sensed the crisis, responded that it will not reduce the live shooting of human models. AI models are just supplements rather than replacements. When each piece of clothing is displayed by one or two human models, AI models are just icing on the cake and can be produced more quickly. More photos.
But that’s less convincing for human models. It is also marginalized models such as people of color and plus-size models who are most vulnerable to the wave of automation, because they are more likely to work at the bottom and have fewer opportunities to choose. According to the “Vogue” business size inclusiveness report, only 0.6% of the various styles in the autumn and winter fashion seasons (New York, London, Paris and Milan) last year were large-sized figures. Melbourne Fashion Week in March was criticized for a lack of models with disabilities. The irony is that the brand ignores the real minority and uses AI models to represent the minority. Richard Lachman, a professor at the Metropolitan University of Toronto, sees the trend as disturbing:
Brands hope to create models that appear to fit the social reality, including all skin tones, body shapes, and ethnicities, but it is an illusion of idealization. AI is not real people, there is really no variety or increased workload.
AI models are also subject to data and algorithm bias, exacerbating certain biases. Vice reporter Chloe Xiang found that the Deep Agency preset produces a blond-haired Caucasian female model. If you want to change the appearance, you must add prompt words such as race and age. This can be extended to predict more tens of thousands of AI pictures, other hair colors and There will be fewer AI female models with skin tones. Experts suggest that AI models should be paired with a diverse behind-the-scenes team, including marginalized human models, for paid shoots to make up for the lack of demographic data in the fashion industry. But there are gray areas that follow: Who really owns the data?
When discussing the various impacts of AI models on the fashion industry, it is always inseparable from the question of Canadian model Sinead Bovell:
Diversity is good, but who benefits and who automates?
The famous economist Joseph Alois Schumpeter’s “creative destruction” may have an answer: the arrival of new inventions, such as electricity and electrical appliances that appeared in the 1880s and 1920s, is like a strong wind blowing, changing industries and the entire economy body. A new technology will eliminate a certain kind of business and industry, and on the other hand, it will also replace the collapsed old technology or industry. Cheaper, faster, and more diverse AI models can certainly speak for minority groups, but they do not benefit the minority groups, at least not directly, and the benefits are more likely to be attributed to the technical elite.
From the perspective of the long river of history, technological progress has indeed benefited mankind, and everyone will eventually benefit. However, during the transition period, many people just paid a silent price, falling into the river and being washed away.
(This article is reproduced with the authorization of Ai Faner; source of the first picture: Levi Strauss)