The inside of the song, licensed for AI training.
XIM Audio licenses the parts a song is made from: real DAW sessions, the stems printed from them, MIDI, synth presets, paired production sets and tutorial sessions, from the WeMakeDanceMusic catalogue, a music production marketplace online since 2012. Eleven datasets, measured and described one by one, about 10,500 hours of audio. Every creator has authorised machine learning in writing and is paid on every licence. No separated stems, no scraped audio.
Every dataset here is a card with the public numbers and anonymised previews behind a gate. The data itself is licensed per set, non-exclusive, on request, with a manifest and a datasheet. Request access on a card and say what you are building; we approve by hand and get in touch.
| Set | What it is |
|---|---|
| stems | 2,817 packs, 61,751 lanes printed from the producer's own session |
| paired-sets | 2,172 songs with session, stems, MIDI, flat and mastered master |
| acapella-instrumental-pairs | 1,486 sample-aligned vocal and instrumental pairs |
| daw-sessions | 12,426 project files, Logic, Ableton, FL Studio, Cubase |
| tutorial-sessions | 383 episodes with timed transcripts, project, stems and masters |
| midi | 338,853 MIDI files |
| presets | 269,305 synth presets |
| loops-one-shots | 708,398 short files from sample packs |
| instrumentals | 3,361 full-length instrumental performances |
| stock-music | 1,925 finished tracks, clean masters |
| marketplace-data | Fourteen years of a music marketplace, de-identified |
Catalogue with every number and the rights per set: ximaudio.com/data. Sample and quote: ximaudio.com/request. Contact: mikas@ximaudio.com. LinkedIn.