Ukrainian Company Creates 290K Shoe Variations with Text-to-Image AI
Hyper-creation in ad campaigns
Ukrainian shoe company KACHOROVSKA says it employed a text-to-image AI latent diffusion model to create “290,186 unique variations of the KACHOROVSKA shoe styles.” That is what the English-language media release said. There may have been a typo, as their provided data suggested that the process took nearly 900 hours. We have sought clarification on this point. However, the news would be significant if only 10% or even 1% of these variations had been produced.
Shoe campaigns might normally consider dozens of images or maybe hundreds but not hundreds of thousands. AI completely transformed the scale of options that KACHOROVSKA had to choose from.
The motivation behind this effort was the company’s new “We Speak Shoes” ad campaign. Generated images were included in four short videos, like the one immediately below.
Hyper-creation with Generative AI
Despite the fact that AI was used to create so many images, Alina Kachorovska, the company’s CEO, suggested that humans were integral to the process. Mood boards were created for each shoe and video while “art direction and creative direction” are required due, in part, to “AI’s variability.”
Most people immediately understand that generative AI models deliver hyper-automation. A single work of art can be created in seconds, while a human might require hours or days to accomplish the same task. However, KACHOROVSKA’s campaign highlights the often-overlooked hyper-creation elements of generative AI.
The AI models can produce a far greater number of image variants than human teams in a short period of time. And the user is never quite sure what will be generated. This leads to many more options to consider in the creative process, and it often means new ideas are discovered that would not have been considered without of the hyper-creation output.
Altered AI talks about a similar value proposition for creating synthetic voices. Its synthetic voice prototyping tool enables users to test a far great number of voice options than traditional synthetic speech models.
You can see more examples on KACHOROVSKA’s Instagram account. The videos are worth a look.
