Project type: Master Project

University: University of Bremen 

Term: April 2005 - April 2006

Supervisors: Prof. Dr. Herzog, Dr. Hermes, Arne Jacobs

Students: Christoph Brachmann, Hashim Chunpir, Silke Gennies, Benjamin Haller, Philipp Kehl, Astrid Paramita Mochtarram, Daniel Möhlmann, Christopher Schultz, Christian Schrumpf, Björn Stolper, Benjamin Walther-Franks

Web-Link: http://www.tzi.de/svp/

Abstract:

Hollywood has long perfected advertising movies to attract audiences. Knowing that every trailer is a piece of art produced by creative humans, we were curious to see how close algorithms could get to this goal. The aim of the SVP project was to automatically generate a movie trailer from an action movie. Therefore, a movie was analysed for several image and sound features (low-level analysis) which were afterwards combined to classify the footage into various categories. Based on this information, the footage of the given movie was automatically selected, edited and mixed with sound elements and on-the-fly created video animation sequences – all according to pre-defined trailer patterns.

Research

Beside film theory and trailer documentation we built our own research pool: an online database we filled with detailed shot-by-shot trailer information. We analyzed 11 action trailers for several aspects such as shot duration, actor appearance, camera position or sound volume. From the resulting pool of data sets we extracted typical trailer patterns. This research formed the basis for modeling our systems trailer knowledge.

The Method

To automatically create a trailer it is necessary to extract the defining features of a movie. Several analyzing operations gathered knowledge about features such as actor appearance, speech, text, sound events and many more. These feature data forms the basis for footage classification and automatic generation. Based on this our generation software creates a fully fledged movie trailer. An underlying trailer knowledge base we fed with results of our research dynamically generates a trailer structure based on semantics. This sequence is enhanced with all kinds of image and sound effects to create a "real trailer feeling". The outcome of the process of fully automatic generation is a video file – the final trailer.

The outcomes of the project were published and presented on several presentations. Sample trailers and further information can be downloaded from the SVP website.