The development of intelligent music production tools has been of growing interest in recent years. Deep learning approaches have been shown as being a highly effective method for approximating individual audio effects. In this work, we propose an end-to-end deep neural network based on the Wave-U-Net to perform automatic mixing of drums. We follow an end-to-end approach where raw audio from the individual drum recordings is the input of the system and the waveform of the stereo mix is the output. We compare the system to existing machine learning approaches to intelligent drum mixing. Through a subjective listening test we explore the performance of these systems when processing various types of drum mixes. We report that the mixes generated by our model are virtually indistinguishable from professional human mixes while also outperforming previous intelligent mixing approaches.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.