For video searching, a set of objects and a relationship between objects in the set of objects is determined in a portion of a script for creating a video. From a collection of video clips, a set of video clips is selected based on matching an object from the set of objects to an object specified in a metadata of a video clip that is selected in the set of video clips. A duration of the relationship in the portion is computed. video analysis is performed on the video clip to determine whether the video clip includes the relationship. When the video clip includes the relationship, a determination is made whether the relationship in the video clip is at least of the duration. When the relationship is at least of the duration, the video clip is selected as a fit for the portion of the script.