Is it real or AI? Fashion industry explores computer generated models


The growing sophistication of artificial intelligence has raised eyebrows in many fields over recent months; so, when Levi Strauss announced it had partnered with a software company to use AI-generated fashion models to showcase its clothes, more than a few heads were turned in the fashion world.

Would software replace the need for flesh and blood models?

In a March news release, Levi’s wrote, “We are planning tests of this technology using AI-generated models to supplement human models, increasing the number and diversity of our models for our products in a sustainable way.”

Amy Gershkoff Bolles, global head of digital and emerging technology strategy at Levi Strauss & Co., said in the news release “While AI will likely never fully replace human models for us, we are excited for the potential capabilities this may afford us for the consumer experience.”

After the initial press release, and subsequent criticism from some in the fashion world, Levi’s and the software firm they are working with,, explained more adamantly that the software is not intended to supplant human fashion models.

In a teleconference interview from Amsterdam, Michael Musandu, founder and CEO of, explained, “I want to be very clear about this like we need real models. We want real models that can create genuine connections with consumers and for brands that are very serious about inclusion efforts they should continue casting models from underrepresented groups.”

But, Musandu said, Lalaland allows fashion brands to extend their diversity reach and marketing toward underrepresented groups. “Instead of a brand showcasing one model and one product, we’re trying to enrich it. You can actually showcase 20-30 different models on one single product with the same efficiency compared to what a traditional photo shoot would be.”

This slideshow requires JavaScript.

How does the AI software work?

“Underneath the hood,” said Lalaland’s founder, “it’s all machine learning… It’s a platform that allows fashion brands to customize their own AI models and the customization means they can select different type of hairstyles, skin complexions, body sizes, ages, poses, whatever they would like to actually configure and that’s then generated and rendered.

“(The resulting models) are not real,” said Musandu. However, “The whole purpose of (Lalaland) is to supplement traditional photography.”

Despite the insistence from Musandu that the software is not intended to supplant fashion models, those who make a living –or dream of making a living as — models are concerned about what happens next. Will the next casting call be handed over to something created with a few clicks of a mouse?

“As a model, I feel unsettled and disappointed about brands shifting their look books to AI models,” said Emmy Lugo, a fashion model from Beacon Falls.

Emmy Lugo of Beacon Falls is a fashion model who is worried about the impact artificial intelligence generated fashion models could have on her future career.

“I believe we have enough technology,” said Tricia Aquilar, a fashion model from Woodbury. “Technology has so many benefits, but it should not replace people.”

The Woodbury resident added,” I always thought that modeling was one of the jobs that a computer couldn’t replace, but clearly I was wrong.”

Even though the argument for AI models such as the ones generated via Lalaland allows for a more diverse representation in catalogues and fashion magazines, Lugo isn’t persuaded.

“For decades, we have strived for change and the industry is more diverse now than ever,” said Lugo. “People are finally starting to feel accepted directly because of real bodies being represented. Why would we put everything we fought for away for even more Photoshop and editing?”

“If they believe that AI models can allow brands to display clothes on a wider variety of body types, then why not use a wider selection of human models,” said Aquilar. “Why does it always have to be a 5’8, size 2, size 6 shoe, with no curves? Why not a 5’3”model, why not also a size 8 model. They have started to include some curvy models, but no other body styles.”

Tricia Aquilar is among the area fashion models worried about a move to use artificial intelligence generated fashion models by apparel companies.

“Ideally,” said Lugo, “I believe brands need models to represent every size they offer. We need to end Photoshop and digital editing to make up these ‘perfect people’ that do not exist.”

“If I were to show up to a casting and was later told I didn’t get the job because they opted for software, I would feel not just upset but concerned for future generations,” said Lugo. “The argument that AI models will allow brands to put clothes on a diversity of body shapes and sizes does not make sense to me. There are plenty of models who come from different backgrounds, shapes, and looks who would love this opportunity. In addition, using AI models would be an inaccurate representation of what the product looks and fits like in reality.”

Laura Kahler’s 13-year-old daughter Lila Van Frachen dreams of being a fashion model. But the Stonington mother said AI could destroy that dream and the dream of many young girls like her middle school student.

Stonington’s Lila Van Frachen, 13, has dreams of becoming a fashion model. But fashion models generated using artificial intelligence technology could make that dream more difficult to achieve.

“My feelings about big brands using AI models makes me sad,” said Kahler. “I think it could replace the need for young models.”

Lila’s mother said, “My daughter enjoys modeling, and it helps her develop communication skills and develop self-confidence. I think our teen generation already has enough time interacting with AI due to social media and electronics especially games. They need more of a human connection, not less.”

Echoing Aquilar and Lugo, Kahler, a school art teacher, said, “I can understand the need for more diverse models; however, it would be so beneficial for the brands to actually hire diverse models instead of creating them out of AI. There are plenty of people who would love to model if given the chance. If these brands advertised what they were looking for I think they would be delighted with the outcome, and I think even more people would respect the brands. It could even result in more sales.”

Rather than creating computer generated fashion models for the sake of diversity, Kahler suggested that brands should “make a documentary about the hiring and interview processes of hiring more diverse models of all ages, sizes and ethnicities.”

Leah Juliett, a Wolcott native and a body positivity advocate, worried that AI could potentially reduce diversity in fashion rather than increase it.

“Let’s first consider why Levi’s would rather use technology to generate diversity instead of doing the internal work to hire, cast, and pay real models with diverse human bodies to showcase their clothes,” said Juliett. “It’s certainly not a cost-saving strategy; Levi’s net worth is over $5 billion. Certainly, they can afford to cast models who physically embody the diversity that the brand wants to convey that they are using AI to reflect.”


Juliett argued, “Levi’s choice to use artificial intelligence to do the brand’s public-facing diversity work instead of making intentional decisions to hire more diverse models may be an easy way to automate representation, but it doesn’t solve the fashion industry’s legitimate diversity problem. By replacing human models with AI-generated models, Levi’s, and any other brand using this practice, is capitalizing on the outward appearance of diversity while simultaneously limiting opportunities and careers for an entire industry of models who are often denied jobs because of their size, race, gender identity, and ability.”

Juliett argued, “Celebrating diversity is good for business. When customers see themselves represented, it drives brand engagement. Even more importantly, prioritizing diversity in your business structure bolsters opportunities for folks who have historically been unwelcome in your industry.”

Musandu acknowledged the lack of diversity in the fashion world; and he said Lalaland was intended to improve the situation.

“We started this company trying to solve a massive problem of people feeling underrepresented online,” said Musandu. “I say this because I’m a person of color and that’s the main reason of never being able to shop at a brand that fully represented models that look like myself.”

The process of generating images via Lalaland acknowledges the lack of diversity in the fashion world, said Musandu.

Although some AI imaging software snags bits and pieces of images found on the internet, Lalaland creates images based on a catalogue of proprietary images provided by the brands. This means real models must be photographed first to create the aggregated images.

“Our approach had to be different because the Internet itself is not diverse enough so if you’re pulling in data from the Internet for an algorithm to generate (images), I think you’re going to have quite a lot of biases and discrepancy in what your output is,” said the CEO.

“For us, that was something that would never work — what it meant is that what we had to do is actually get our own proprietary data,” said Musandu. “That means that we’ve actually created new monetization opportunities and revenue incentives structures for BIPOP (Black, Indigenous, and People of Color) community groups to license their data or sell their data. This is new proprietary data that (didn’t exist) on the Internet.”

To use Lalaland, brands will need to supplement traditional photography to provide the software with the data it needs to create the AI images “just to ensure that there’s more representation being shown,” said Musandu.

For many brands, said Musandu, it’s not just a matter of hiring more models to improve diversity. And that’s where Lalaland helps brands.

“Typically brands, before they have to launch a collection online, they have to hire real models… venues, photographers, hair stylists, makeup artists,” said Musandu. Since the brand must do this several times a year, the CEO said, “It’s almost infeasible for them to showcase, let’s say 20 models or one product, without having to exponentially increase what the product price would be… A lot of brands would only have 20 models for the whole year.”

Using Lalaland and AI, Musandu said brands could have 200-300 models represented in their clothing.

Brands can supplement a traditional photo shoot “in a couple of minutes as soon as you have your product item instead of waiting weeks or months… going from your sample to having an on-model image… That efficiency allows you then to amplify the (number) of models and (the amount of) content that you can produce.”

Musandu also said, “We’re also helping level the playing field for small emerging brands… (that) don’t even have budgets to have photo shoots… What we do is help the underdog… They can compete head-to-head with these global brands because they have the same quality models in a much more cost-effective way.”

Artificial diversity through artificial intelligence, however, will not sway Juliett when it comes to parting ways with any hard-earned cash when shopping for clothes.

“As a consumer, I’m not going to engage with or celebrate a brand that used artificial intelligence to make me feel represented in its effort to sell me a product,” said the body positivity advocate. “I am going to celebrate — and consume –a brand that puts its money where its mouth is and has ethical and diversity-forward business practices. A brand that pays its staff equal wages, makes clothing for all sizes and doesn’t gate-keep its products for bodies that are deemed the most profitable, marketable, or consumable.”