Stem separation is a practical on-ramp to building personal sample banks from any released track

For a live coder or sampler-oriented producer, neural stem separation transforms any commercial or freely licensed recording into a source of isolated sounds: a clean kick pattern, isolated bass line, a capella vocals, or textured accompaniment. These can be chopped, pitched, time-stretched and fed directly to Strudel, Tidal, or a hardware sampler. Demucs is the free, MIT-licensed entry point: no API key, no subscription. The key limitation: separating copyrighted tracks for distribution remains legally uncertain even when separated stems are not reproduced verbatim — check the license of the source material.

Examples

Workflow: demucs --two-stems=vocals track.mp3 → import vocals.wav into Strudel as a sample → chop and trigger with pattern language.

Alternative: use htdemucs_6s to extract isolated guitar → use as a one-shot in live set.

Assessment

Describe a complete workflow from downloading a CC-licensed track to loading an isolated drum stem into a Strudel pattern. What legal check do you do first?