Unsupervised Scouting and Layout for Storyboarding in Movie Pre-production

We introduce a system that analyses film content to provide expert camera pose recommendations to cinematographers, so they can rapidly experiment with shots in a virtual space. Shot experimentation to produce storyboards is a crucial preparation step prior to filming, with the time and skill required to create expert storyboards high. Additionally, the convergence between the desired aesthetic and what the physical environment allows is traditionally explored by scouting on location, which can also be costly. We hypothesise that autonomous scouting in the virtual world by moving cameras, actors and lights to achieve specific composition traits is efficient and supports creativity. Our system automates aspects of scouting, storyboarding, and staging without being directed by limiting rule-based cinematography techniques, and instead by data-driven unsupervised clustering.