AlbumDB: A multitrack dataset of a ten-song album with stereo and immersive 7.1.4 masters

For a detailed description of the dataset, please see:McKenzie, T., Moliner, E., and Wright, A. (2026). A multitrack dataset of a ten-song album with stereo and immersive 7.1.4 masters. In proc. AES 160th Convention, May, Copenhagen. This multitrack dataset is intended to support music production research and education, including machine learning techniques such as automatic mixing and source separation. The dataset comprises a cohesive 10-song indie album titled `Extraordinary' by Tom McKenzie (indie rock/folk), with separate stems for individual instruments, such that each song has between 13 and 35 individual tracks (stems). This album-format dataset enables studies of mix consistency across a thematically aligned collection of songs, as well as stereo upmixing to immersive formats, and contains far more stems per song than traditional four-stem datasets. All audio is supplied as 16‑bit, 44.1 kHz PCM WAV. Stem Mixing VariationsFor each song, three versions of every stem are provided: Raw: the original recorded tracks with no mixing or processing; the only changes are temporal edits and removal of silences. Mixed-Dry: each stem after human mixing with level balancing, panning, EQ, and compression, but with all time-based effects (reverberation, delay) omitted. These stems are used (along with separate immersive time-based processing effects) to make the immersive 7.1.4 masters. Mixed: as mixed for the dry stems, but with the full mix processing (reverberation, delay, etc.). These stems are used to make the final stereo masters. MastersFor each song, two final mixes are provided: Stereo: the standard two‑channel stereo masters 7.1.4: 12-channel immersive 7.1.4 masters, made using the `Mixed-Dry' stems. To preserve dynamic range and avoid overcompression of the centre channel, the 7.1.4 masters were rendered at an RMS level 8 dB lower than the stereo masters. For level‑matched comparisons, users should either increase the immersive masters by 8 dB or reduce the stereo masters by 8 dB. Example Usage - Automatic MixingTo illustrate an example usage of the dataset, two songs were mixed using the MEGAMI automatic mixing model: Song 01 - Extraordinary, and Song 10 - It Ain’t Easy (To Mend A Heart). These are included here. Changelog V 1.0 - Initial version