Stewart's directorial debut, "Come Swim," is on the short-film program this week at the Sundance Film Festival, and it's a trippy meditation that blends artistic vision with one of today's most advanced computing techniques.
To lay the groundwork for the 17-minute movie, Stewart last week published a research paper that details how she got her impressionistic visual effects -- through the use of neural networks, an aspect of artificial intelligence in which a collective of computers tries to emulate the workings of the human brain.
The specific technique, called Neural Style Transfer, uses neural networks to "artistically redraw an image" in the manner of an impressionistic painting, made by Stewart herself, of "a man rousing from sleep," the paper says.
After adding blocks of color and texture to the transfer, the effect is both dreamlike and surreal. The process is used throughout the film.
"Neural Style Transfer requires many creative iterations when trying to work towards a specific look for a shot," the paper states. "The parameter space is very large -- careful, structured guidance through this in collaboration with the creative leads on the project is needed to steer it towards a repeatable, visually pleasing result."
The paper, entitled "Bringing Impressionism to Life with Neural Style Transfer in 'Come Swim'," doesn't tell how many computers were pressed into service for the film. It does, though, cite an Oxford University-affiliated neural network (for an evaluation that "provided the right balance between execution time and the quality of result we desired") and the Amazon EC2 cloud computing service (for graphics-processing resources that would "get us to the needed level of quality" for production).
The paper was authored by Stewart; by the film's producer, David Shapiro; and special effects engineer Bhautik J. Joshi. It was published Jan. 18 in arXiv, an online repository for non-peer-reviewed work.
Solving for XX: The industry seeks to overcome outdated ideas about "women in tech." Take a look here.
Crowd Control: A crowdsourced science fiction novel written by CNET readers. Read it here.