XIM Mix · measured 25 Oct 2026
Training data for AI mixing and mastering
A mixing model needs the decisions, not only the result; the printed lanes carry the insert chains, the parsed project carries gain, pan, sends and plugin lists, and the flat and mastered masters are the targets.
Each song comes with
- every lane as printed through its inserts (wet)
- the flat master and, where the producer made one, the mastered master
- the project file: Ableton sets rewritten with names scrubbed; Logic, FL Studio and Cubase projects as saved, with the names inside them the producer's own
- project.json from the parser: tracks, plugins, samples, tempo, key, arrangement, every name scrubbed
- on order: the same lanes printed dry and every device parameter exported
On order: dry lanes (effects bypassed); device parameter export per lane.
Built from
- DAW sessions The producer's project file with its audio and MIDI, as it was when the track was finished.
- 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.
How it ships
Readiness: ships within 10 working days of payment. 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
Sold project files under terms that name AI training. Read the rights page.