Summary

CONCEPT

Interwoven

with Generative AI.

“You can easily create designs with generative AI.” — Such phrases have become increasingly common in recent years.

But can AI truly engage in acts of creation with designers? This fundamental question marks the starting point of our project.

In this experimental initiative, we train AI to learn the formal characteristics unique to the designer, guiding it through prompts that reflect personal tastes and creative directions, to explore a process of genuine co-creation between human and machine.

Through this collaboration, the project also seeks to horizontally expand the very definition and role of what it means to be a designer.

“To create images without names” — UNTITLED IMAGE CLUB.

Here, we present the documentation of this ongoing exploration.

image.png

Step

  1. The designer’s knowledge, formal tendencies, and recurring visual expressions are distilled into a unique identity and trained into a LoRA model. Prompts are then prepared for generating object forms.
  2. Using this LoRA model and the crafted prompts, images are generated—producing object visuals on screen.
  3. From the diverse range of generated images, those that most closely embody WD’s distinctive sensibility are selected, and several are subsequently transformed into three-dimensional forms.

image.png

LoRA

This process involves transforming WD’s creative output into a compact AI model (LoRA) and training it to internalize the studio’s distinctive formal language—thereby bringing the generated results closer to the essence of WD.

Each LoRA was trained on approximately twenty images. In selecting these images, a deliberate balance was maintained between full-view compositions and detailed close-ups, ensuring that the dataset encompassed a wide range of visual characteristics and expressive nuances.