Picking the Editors Pick
In the media industry, trailers for video content such as movies and series are a core medium for raising awareness, branding and enticing audiences to watch. At RTL, a team of a dozen promo editors produce trailers for hours of video every day. Making trailers for movies and series is an expensive and labor-intensive job. Video content must be watched completely, after which the most interesting scenes are manually selected and finally edited into a short trailer.
A 90 minute movie trailer normally only lasts 10 to 30 seconds, but it takes half a day to make. One problem is that editors have to spend most of their time manually parsing video data. A second limitation is that only a one-size-fits-all trailer without personalisation can be produced for each video.
This project goes beyond one-size-fits-all trailer production and focuses on personalised video trailer editing. The goal is to deliver a Key Enabling Methodology (KEM) that enriches the editing process of creative professionals. As a basis, we will first develop a Key Enabling Technology (KET), namely a neural network that learns to score video scenes for different genres based on their likelihood of being usable in a trailer. Using this technology, we will develop an interactive interface, where editors will automatically see the best-ranked scenes in different genres, alleviating the need for tedious manual data inspection.
Project budget €126,800, with a €105,600 PPS surcharge being used.