Demos / Music
AMEN.
Cut a jungle dub with a tiny trained composer.
Tap CUT A DUB in the studio and let it compose. Tap Play if audio does not start automatically. Change Variation for the next take, or select an earlier dub to play it again.
Inside the composer
Music from
learned records.
AMEN learned from 6.2 hours of music. A persistent context selects a sparse set of records to predict the next musical event.
This is the classic Cadence 0.11 record-memory design: 128 context coordinates, 8,192 record cells and 48 active records per reading. A supplied instrument and rhythm representation turns its event predictions into drums and bass.
The page composes the track and renders its audio before playback. The animated activity is the recorded generation trace. Learning is disabled in this browser checkpoint.