Artificial intelligence chooses the soundtrack for pictures and movies
Picture Categorization for Suggesting Soundtracks (PICASSO) is new software that analyzes existing films to suggest soundtracks for your picture albums or movies. A high-speed car chase might be accompanied by a techno track, while a walk through the countryside might be set to classical music.
The system’s creators, Aleksandar Stupar and Sebastian Michel at Saarland University in Saarbrücken, Germany have programmed it to suggest the top ten tracks for slides in a photo album or movies by matching them to stills in existing films. The current version selects between just a few hundred copyright-free sound tracks, but Stupar and Michel plan to create an iPhone version that would allow the system to automatically suggest tracks from the user’s MP3 collection.
Michael Jewell, a multimedia researcher at the University of Southampton, UK, says there is a need for this kind of software, but images alone aren’t enough to determine the mood of a soundtrack. “It’s difficult to say if a picture of a car is from a frenetic car race or a failed driving lesson,” he says.