Pick any song as the seed. Sensei scores everything you own against it, clears the next track, and renders the seam: key-matched, phrase-aligned, FX-racked. Then it does it again. Your library, playing itself like a set.
↑ hit skip and watch the engine pull the next seam in. That's the steering model
Any track in your library is a valid first move. Sensei builds the run outward from it: harmonic neighbors, matched energy, phrase-clean exits. A 2,000-track library is a few hundred hours of sets it can construct.
Don't like where it's heading? Skip pulls the next seam in immediately. Want the room warmer or cooler? Nudge the energy target and the candidate ranking re-sorts around it. You steer; it mixes.
This isn't the 6-second volume fade your player app calls "mixing." Every seam is planned: lows swapped at the right bar, filter sweeps, an artist-styled FX rack armed around the swap point. The same engine working DJs use in Prep mode.
Listen mode plays your files (mp3, wav, aiff, flac). No streaming subscription, no licensing middleman, no "this track is no longer available." If it's on your disk, it's in the set.
Listen mode works with local files only. It does not and will not plug into streaming services. That's not a missing feature; it's the design. Your crates are the catalog.