XIM Transcribe · measured 25 Oct 2026
Training data for music transcription
Audio-to-MIDI needs the notes that actually played the audio, on the same clock; here the MIDI is the part the producer wrote and the lane beside it is what it played.
Each song comes with
- every lane as a WAV
- the MIDI files of the session, names scrubbed
- midi_pairs.json: which MIDI file played which lane, with the method and score of the match
- mix.wav and data.json
Built from
- Paired production sets One song, everything that made it: the producer's session, every lane printed from it, the MIDI it played, and the flat and the mastered master.
- MIDI files Note data from sessions and MIDI packs: chords, leads, basslines, drums.
How it ships
Readiness: ships within 10 working days of payment; the pairing check in the datasheet runs on every delivery. Delivery to a private S3 prefix or your bucket, with a manifest, checksums, creator-level splits and a datasheet.
Loader: one folder per song with data.json.
Rights
Every creator in this set accepted terms that name AI training. Every part comes from the creator's own session, and the manifest says which session it came from. Read the rights page.