Algorithmically Generated “Average” Faces From Films

July 19, 2013 | Julia Dawidowicz

Movies consist of countless frames of countless faces in countless situations. How can a movie’s “identity” be summed up into one image? This is what Seoul-based artist duo Shinseungback Kimyonghun sought to find out with their project Portraita series of algorithmically-generated composites of faces from films like Kill Bill: Vol. 1The MatrixOldboy and Taxi Driver.

The artists used facial recognition software to detect faces every 24 frames of a film, and then layered all these images to created an average “face of the movie.” The resulting faces have an eerie deja vu-like quality to them. Perhaps unsurprisingly, the images all look quite the same (except for the Oldboy composite, which is distinctly, well, Asian), like androgynous ghosts submerged in a milky liquid.

“Some faces seem to reflect the original movie’s identity better, while others look obscure,” Shinseungback Kimyonghun tells Wired UK. “But the ambiguity itself might also convey information about the movie: many people featured in it, the cinematographer shot in various angles and so on. Some people will find some of the images creepy because of the look caused by averaging process, but we think that’s a part of the work.”