Marc-André Weibezahn

Beyond Pastiche

Generated Images, Painting, 2017

The technique of style transfers – rendering one image in the style of another one, e.g. digitally painting "Mona Lisa" in the style of Monet – got much attention during 2016. Breakthroughs in Deep Learning made it possible to create digital visual pastiches with much more fidelity while working faster and more flexible than previous approaches. To the public it shortly seemed as if we are one step closer to computer programs that can be "creative".

The results were several websites, mobile apps and source code releases which enable everyone to create such digital pastiches. Most users stick to a quite narrow area: they take a picture (mostly portraits) and use a well known artistic style. Results look very similar after a while.

"Beyond Pastiche" goes further and explores the combinations of unusual inputs for this technique. Images of meat, litter, icons of pop culture and consumption are applied to different motives to create new contexts.

One further step was then to paint one of the results on actual canvas.

tree
"My Image", painting, acrylic on canvas, 90 by 90 cm
tree
Style tranfer: Tree + litter
sneaker + care bears
Sneaker + Care Bears
lion + cigarettes
Lion + Cigarettes
lambolicious
Lamborghini + Meat
scream
Self portrait neutral + self portrait screaming

Works

Raeliteis
emofi
Occupational Therapy
Beyond Pastiche
Meditations on Time
Zirkel
Phacemix
to Bed