A system, method and computer product for training a neural network system. The method comprises applying an audio signal to the neural network system, the audio signal including a vocal component and a non-vocal component. The method also comprises comparing an output of the neural network system to a target signal, and adjusting at least one parameter of the neural network system to reduce a result of the comparing, for training the neural network system to estimate one of the vocal component and the non-vocal component. In one example embodiment, the system comprises a U-Net architecture. After training, the system can estimate vocal or instrumental components of an audio signal, depending on which type of component the system is trained to estimate.

A system, method and computer product for training a neural network system. The method comprises applying an audio signal to the neural network system, the audio signal including a vocal component and a non-vocal component. The method also comprises comparing an output of the neural network system to a target signal, and adjusting at least one parameter of the neural network system to reduce a result of the comparing, for training the neural network system to estimate one of the vocal component and the non-vocal component. In one example embodiment, the system comprises a U-Net architecture. After training, the system can estimate vocal or instrumental components of an audio signal, depending on which type of component the system is trained to estimate.

Musical catalog amplification services that leverage or deploy a computer-based musical composition system are described. The computer-based musical composition system employs algorithms and, optionally, artificial intelligence to generate new music based on analyses of existing music. The new music may be wholly distinctive from, or may include musical variations of, the existing music. Rights in the new music generated by the computer-based musical composition system are granted to the rights holder(s) of the existing music. In this way, the musical catalog(s) of the rights holder(s) is/are amplified to include additional music assets. The computer-based musical composition system may be tuned so that the new music sounds more like, or less like, the existing music of the rights holder(s). Revenues generated from the new music are shared between the musical catalog amplification service provider and the rights holder(s).

Computer-based systems, devices, and methods for assigning mood labels to musical compositions are described. A mood classifier is trained based on mood-labeled musically-coherent segments of musical compositions and subsequently applied to automatically assign mood labels to musically-coherent segments of musical compositions. In both cases, the musically-coherent segments are generated using automated segmentation algorithms.

Computer-based systems, devices, and methods for generating variations of musical compositions are described. Musical compositions stored in digital media include one or more music data object(s) that encode notes. A first set of notes is characterized and a transformation is applied to replace at least one note in the first set of notes with at least one note in a second set of notes. The transformation may explore or call upon the full range of musical notes available without being constrained by conventions of musicality and harmony. For each particular note in the second set of notes that replaces a note in the first set of notes, whether the particular note is in musical harmony with other notes in the music data object is separately assessed and, if not, the particular note is adjusted to bring it into musical harmony with other notes in the music data object.

Computer-based systems, devices, and methods for generating musical compositions are described. A population of musical compositions stored in digital media are each segmented to produce abridged samples. The samples are analyzed to identify “parent” compositions that best exhibit or evoke a particular desired quality. The parent compositions are cross-bred to generate a set of child compositions which are similarly segmented and analyzed. The child compositions that best exhibit or evoke the particular desired quality are re-cast as parent compositions from which another generation of child compositions are bred. Mutations in the form of musical variations are inserted in at least some iterations and the process is repeated until at least one child composition that satisfies some exit criterion is returned.

Methods and systems for artificially generating media streams are provided. Textual information, rhythm information and voice characteristics may be received. It may be determined that a first portion of the textual information corresponds to a first portion of the rhythm information and that a second portion of the textual information corresponds to a second portion of the rhythm information. Audio stream may be generated based on the textual information, the rhythm information and the voice characteristics. A first portion of the audio stream may include a vocal expression of the first portion of the textual information in a voice corresponding to the voice characteristics and according to the first portion of the rhythm information, and a second portion may include a vocal expression of the second portion of the textual information in the voice corresponding to the voice characteristics and according to the second portion of the rhythm information.

Computer-based systems, devices, and methods for generating musical note sequences are described. One or more musical composition(s) stored in digital media include one or more data object(s) that encode notes and/or note sequences. At least one note sequence is processed to form a time-ordered sequence of parallel notes, which is analyzed to determine a k-back probability transition matrix for the at least one note sequence. An attribute, such as a style, of the at least one note sequence is thus encoded and used to generate new note sequences that embody a similar attribute or style. In some implementations, the at least one note sequence may include a concatenated set of note sequences representative of a particular library of musical compositions.

In implementations of searching for music, a music search system can receive a music search request that includes a music file including music content. The music search system can also receive a selected musical attribute from a plurality of musical attributes. The music search system includes a music search application that can generate musical features of the music content, where a respective one or more of the musical features correspond to a respective one of the musical attributes. The music search application can then compare the musical features that correspond to the selected musical attribute to audio features of audio files, and determine similar audio files to the music file based on the comparison of the musical features to the audio features of the audio files.

A device for processing audio signals, including: first and second input units providing first and second input signals of first and second audio tracks, a decomposition unit to decompose the first input audio signal to obtain a plurality of decomposed signals, a playback unit configured to start playback of a first output signal obtained from recombining at least a first decomposed signal at a first volume level with a second decomposed signal at a second volume level, such that the first output signal substantially equals the first input signal, and a transition unit for performing a transition between playback of the first output signal and playback of a second output signal obtained from the second input signal. The transition unit has a volume control section adapted for reducing the first and second volume levels according to first and second transition functions.