'NYC Flow' Uses a Neural Network Algorithm to Paint a Scene

By: Joey Haar - Jul 28, 2016
References: vimeo & digitaltrends
'NYC Flow' combines cutting edge computing power and a PhD-level algorithm with beautiful painting and haunting slow-motion scenes to create a unique piece of art. The NYC Flow video was created by digital artist Danil Krivoruchko in collaboration with Manuel Ruder and a team from the University of Freiburg in Germany. The clip shows several shots of average scenes from in and around Manhattan but colors them in a style that makes it look like a moving watercolor painting.

Turning live images into painting is nothing new, as the Prisma app allows users to add the effect to their iPhone photos, however the app does not yet support video. NYC Flow's algorithm is able to take as input any painting's style, be it graffiti or, as is the case here, dreamy Impressionism, and apply that to every frame of a video. The result is a moving image that could very well represent a brand new genre of art: the moving painting.