How the vocabulary becomes Strudel

strudel-bench songs are written with a small musical vocabulary: eight layers (drums, bass, melody, pad, fx, sample, perc, raw), twelve axes from 0 to 1 where 0.5 is each layer's baseline, descriptors that move several axes at once, modifiers that scale them, and harmony words that pick keys and progressions. A song is a list of sections, each with its own layer settings, and an edit request names the section it is about. An axis can also be a signal that moves through its section instead of a constant. Drums, bass and perc can also write their rhythm directly as a grid of hits, accents, ghosts and rests. A continuous axis can carry a named movement — wobble, drift, pulse, swell — instead of a plain ramp. And a whole section can be reshaped by an arrangement verb like breakdown, lift, strip or halftime.

Why write at this level? Plain Strudel says what to play: a sample name, a note list, a filter cutoff in hertz. The vocabulary says what it should sound like, and the layers turn that into the right pattern calls for their instrument. "Darker" is one word instead of a cutoff, a gain change and a sample swap across three layers, so an edit request stays small and a change touches only the section it names. Because every axis has the same range and baseline, "a bit brighter" means the same thing for a pad as for a bass line, and a request can be checked by rendering before and after rather than by ear alone.

Every entry below is playable and editable. Press a variant, listen, then open the expanded Strudel: the top pane is the high-level language you write, the bottom pane is the plain Strudel it becomes, regenerated as you type. Only one example plays at a time.

Now go try it yourself

Every example above was a sentence before it was a pattern. Head over to Compose, open a song, and ask for it to be darker, punchier, sparser, or more dreamy. The vocabulary does the arithmetic; you do the listening. And if it sounds wrong, that is not a bug, that is a request for the next word.