Fashion Tech News symbol
Fashion Tech News logo
2024.03.22

The Japan-Originated 'AI Model': How Does It Help Apparel Companies' DX? – AI model INC.

In recent years, amid great attention being focused on generative AI, AI model INC., a company that generates AI models using Japanese AI technology and supports DX, is being talked about. They assert that utilizing AI models in areas such as apparel e-commerce and photo shoots, including so-called 'Sasage-gyomu' tasks, can improve sales and reduce costs.
For this piece, we've spoken to Yuuki Nakayama, the CTO of the company, about everything from the origins of their AI modeling business, the details of their initiatives, to their future plans.

Challenges that Arose from a Surge in E-Commerce

Could you tell us about how your company started this initiative to generate AI models/talents with AI technology and support DX?
Just before the onset of the coronavirus, apparel e-commerce sales were surging. However, the sudden rush to adopt e-commerce led to a problem. So-called 'Sasage-gyomu' tasks (shooting, sizing, creating product details for e-commerce websites, abbreviated in Japanese as 'ささげ業務') became extremely labor-intensive.
Brands that utilize models for product detail photography often faced issues like being unable to shoot due to cost constraints when minor batch photography was necessary, or missing posting deadlines due to not being able to respond to the product's lifecycle. Furthermore, there are still many brands that, unable to afford model photography, continue to only post images of their products on e-commerce sites. These were some challenges that arose because of e-commerce's growth.
Contemplating on whether these problems could be solved by using cutting-edge technology, we started AI research and development around 2018, before generative AI started making headlines. We then established 'AI model INC.' in 2020 and launched our service in August 2022.
The biggest difference between our company and other AI companies is that we prioritize the market rather than the technology. Our representative Taniguchi, also the representative of a creative production company, has been dealing with various tasks like constructing websites and e-commerce sites, producing LookBooks and catalogs, and doing art direction for magazines for major apparel companies.
He used to handle photography for 'Sasage-gyomu' tasks, so the service was born as a result of considering the most suitable technology and how to provide it after understanding the business practices of the fashion industry.

Generating AI Models with an In-House Engine

How did you build the AI technology that generates models?
Technically, we're not using Midjourney or Stable Diffusion, but generating the AI models entirely with our in-house engine.
We create dedicated AI models for the companies that request our service, and we also handle creative production.
There must be certain difficulties in generating AI models that client companies request. How do you see it?
As you say, it's indeed the most challenging point. For instance, in human model auditions, there are several judgement factors such as character, track record, image, background, etc, so even if there are multiple people making the decision, all often agree on 'this one is good'. However, with AI, the only component is the image, which makes the decision difficult.
Therefore, our in-house creative directors understand the client company's intentions, screen the AI generated in-house, and then propose.
1 / 2 pages
Share Article