摘要 |
Provided are, among other things, systems, methods, software programs and techniques for generating an audio thumbnail for a musical piece, in which locations of different repeating segments within an input musical piece are identified. In addition, a singing segment within the musical piece is detected based on a calculated singing metric that indicates when singing is present. A thumbnail criterion that defines a location of an audio thumbnail by reference to generic musical structure is obtained, and a segment of the musical piece is selected as the audio thumbnail based on the thumbnail criterion, together with the location of at least one of the plural different repeating segments and a location of at least a portion of the singing segment. The audio thumbnail is then played and/or stored for future playing. |