Music Data for AI Training: 1,334 Licensed Acapella and Instrumental Pairs from the Same Session

Acapella and instrumental pairs is the clean ground truth for separation: 1,334 pairs, 2,668 WAV files, about 146 hours of audio, counted on 14 October 2026. What it is Each pair is the vocal and the instrumental of one finished song, printed from the same session. Both files share one clock, so the vocal and the instrumental sum to the mix. No separation model touched these files. 24-bit stereo WAV, with BPM, key and genre set by the producer who made each track. The pairs cover house, techno, trance, drum and bass, pop dance and other electronic genres from working producers. What is new 380 of the acapellas were printed in October 2026 from the producers' own sessions, 277 from one remake producer and 103 from another. 140 of the pairs are not on the public store yet, and about 60 more pairs are in the pipeline. Beyond the pairs, the catalogue holds 2,238 acapella covers in total; 893 of them have no instrumental counterpart and are available as vocal-only data. A further 117 house originals with vocal stems are identified for the next batch. Why it matters Public multitrack sets for separation research are small and non-commercial. This set is commercially licensed, measured, and built the way a separation model wants its targets: the real vocal print and the real instrumental print of the same session, not a mix and a model's guess. Many of the songs also have their full stems and the DAW session in the sets beside this one. Where it comes from The songs were sold on WeM