Synthetic Feathers

“Synthetic Feather” is an algorithmically made knitwear piece created to accompany the HATRA 2020 Autumn Winter Collection “STUDY SKINS”.

From the myriad of images that exist in cyberspace, the GAN (Generative Adversarial Network) algorithm generates imaginary birds. The generated images are converted into data that can be output by a knitting machine in collaboration with a textile craftsman, and the patterns that symbolize the theme of this season “Study Skins” are saved as sweaters.

About the Project

The knitwear “Synthetic Feather” features the appearance of “imaginary birds” generated from a GAN (a type of machine learning known as Generative Adversarial Network) based on 2 million images. These are entities born within the information environment of today where vast amounts of images drift, and may serve as a guide for us to contemplate new ecosystems dreamt by machines.

This work represents a continued exploration of new digital expressions that deviate from optimization, following on from AUBIK, and was presented as a collaborative work between Synflux and HATRA.

Study Skins Collection (2020 AW)

HATRA’s Autumn/Winter 2020 collection “Study Skins” explores the concept of taxidermy and preservation through digital transformation. The Synthetic Feather pieces serve as preserved artifacts of imaginary birds found in the digital realm.

Synthetic Feather - Study Skins Collection

Bird Variants

The project explores multiple bird species generated through the GAN process, each represented as a unique knit pattern: Owl, Peacock, Penguin, and Toki (Japanese crested ibis). These imaginary birds exist only as data, brought to physical form through textile manufacturing.

Ornithopter Collection (2021 SS)

The Spring/Summer 2021 collection “Ornithopter” continued the exploration, presenting evolved versions of the Synthetic Feather pieces with refined knitting techniques and more complex patterns.

Synthetic Feather - Ornithopter Collection

Technical Process

Data Collection: Gather 2+ million images of birds from the internet
GAN Training: Train a Generative Adversarial Network to generate new, imaginary bird species
Pattern Generation: Convert AI-generated images into knitting machine-readable data
Textile Production: Collaborate with textile craftsmen to produce physical garments

The process transforms digital data into tangible form, creating a bridge between computational imagination and material reality.

Art & New Ecology Exhibition (2022)

The work was presented at “Art & New Ecology” exhibition at Tokyo University of the Arts Museum, exploring themes of technology, nature, and new forms of creativity emerging from their intersection. The exhibition considered how machines, human hands, and natural environments might guide new forms of creation in an era of accelerating technological evolution.