Robbie Barrat x Balenciaga


 ︎︎︎ Acne Studios
In the summer of 2018, the artist and programmer Robbie Barrat wrote a short post to his Twitter feed: “I’m doing something with fashion and AI but I don’t know what yet.” Later that day, he identified that he would use Facebook’s DensePose dataset for training and testing. Shortly after, Barrat posted an image of the Balenciaga website noting that it already looked like a readymade dataset, and then posted to say that he had written a short program—“only 4 lines long”—to scrape all the images from Balenciaga’s online lookbooks.

In the days that followed, Barrat posted uncanny images and astute observations on the capacity of an AI neural network to learn the codes of contemporary fashion. He identified anomalous designs or unexpected combinations—the “super high shoulders/collar,” a “button up shirt windbreaker combo,” or a “black jean and sweater all in one piece.” Symmetrical design is fundamental to fashion, but Barrat recognized remarkable asymmetries in colour and construction. In Barrat’s images, fabrics take on unexpected texture, dissonant colour combinations and a general instability. Accessories often appear as scraps of fabric held by the model, huge belts or bags fused to legs.

Barrat’s creative practice is equal parts research and production. His extensive background in programming and AI research is matched by his commitment to an interdisciplinary creative practice that reaches across genre. Taking many of his aesthetic cues from the legacy of Surrealism, Barrat seeks a new space for creativity between the learned and the unexpected. 

Drawing on the extensive online documentation (lookbooks, catalogues and marketing campaigns) produced by the renowned Paris-based fashion house, Barrat created an archive of images that proposed a new identity for Balenciaga in the age of artificial intelligence.

Using a generative adversarial network (GAN), Barrat was able to closely monitor the machine learning process he had designed, and identify selected images from within the GAN’s latent space (all the possible images laid out in  highly dimensional space) for consideration. It is this aptly named latent space—hidden, imperceptible, inchoate—that is so often utilized by contemporary artists and designers who work with neural networks.

Text To Speech


Robbie Barrat, Robbie Barrat x Balenciaga, Images: Twitter @videodrome, July 16 – August 21, 2018