This paper presents a multitrack dataset intended to support research and education in music production. The dataset comprises a cohesive 10-song indie album (indie rock/folk) with separate stems for individual instruments, totalling between 13 and 35 individual stems per song. For each song, stems are provided in three variants: raw unprocessed stems, mixed stems but without reverberation or delay effects, and fully mixed stems. Additionally, two master formats are provided: stereo and immersive 7.1.4. This album-format dataset enables studies on mix
consistency across a thematically aligned collection of songs, as well as stereo upmixing to immersive formats, and contains many more stems per song than traditional four-stem datasets. Finally, to illustrate an example research usage of the dataset, the MEGAMI automatic mixing model is used to mix two songs. The results show that many of the MEGAMI mixing decisions are similar to those of the human mixes. The dataset is made open-access and free to download.
(See document for exact affiliation information.)
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 Join the AES. If you need to check your member status, login to the Member Portal.
