Researchers have taught AI to automatically generate movie trailers

Agree, sometimes in the cinema, trailers of future films that are played before the show turn out to be more interesting than the next sequel to the next sequel, which you came to watch because of the trailer you saw earlier. Often (too much often, according to the editorial board gg) it so happens that the film simply does not live up to our expectations, provoked by a quality trailer.

Creating such exciting trailers is a whole science. To edit a good trailer even for a bad movie, you need to be at least a good psychologist. Correction: need It was be a good psychologist.

Researchers at the University of Edinburgh say they have developed an AI model based on a pair of neural networks that is capable of delivering a suitable, emotionally engaging trailer for any movie. The team has already used this system to create over 40 trailers for existing films, and Amazon Turk employees generally preferred the AI-generated trailers to the official ones.

The task is not as simple as it might seem. A word to the authors:

To automatically generate trailers, we need to perform low-level tasks such as identifying a person, recognizing actions and predicting moods, as well as higher-level tasks such as understanding the relationships between events and their causation, as well as inferences about characters and their actions.

The researchers were able to achieve the result using two separate neural networks. The first processes the video and audio footage of the film to find interesting scenes. And the second, in fact, is the arbiter of what may be of interest to the viewer. She scans the text version of the movie - like a script - and uses natural language processing to find important and emotional moments.

The finished model generates new trailers using "movie insight" based on how neural networks process input.

A source: techxplore