High-Level Control of Drum Track Generation Using Learned Patterns of Rhythmic Interaction
This work addresses the need for better user control in computational music generation, particularly for drum tracks, though it is incremental as it builds on existing deep learning approaches.
The authors tackled the problem of user control in conditional music generation by proposing a model for kick drum track generation that uses learned relational codes to encode desired interactions with existing musical material, enabling the creation of varied and musically plausible tracks and pattern transfer between songs.
Spurred by the potential of deep learning, computational music generation has gained renewed academic interest. A crucial issue in music generation is that of user control, especially in scenarios where the music generation process is conditioned on existing musical material. Here we propose a model for conditional kick drum track generation that takes existing musical material as input, in addition to a low-dimensional code that encodes the desired relation between the existing material and the new material to be generated. These relational codes are learned in an unsupervised manner from a music dataset. We show that codes can be sampled to create a variety of musically plausible kick drum tracks and that the model can be used to transfer kick drum patterns from one song to another. Lastly, we demonstrate that the learned codes are largely invariant to tempo and time-shift.