Acid Sequence

Rhythm programming

Autechre-Style Beats: Why Random Notes Aren't the Answer

Hit Randomise on a drum rack and you get notes nobody planned. The rhythms people admire in Autechre's music, as far as the duo's own interviews describe them, owe more to rules chosen on purpose and judgement about what to keep.

Below is an original experiment for any DAW: one five-step click against a 16-step bar. It takes five bars to come round again, and one parameter changes it. It is our example, not their method.

Acid Sequence Editorial · Published · Opinion

Randomise is not a rhythm

Hit Randomise on a drum rack and you get a pile of notes nobody planned. Some of it sounds busy, some of it broken, and the pattern can't tell you which is which, because no relationship in it was ever meant. Beats that wander and still feel intentional come from somewhere else: a few rules you can state, left running long enough to surprise you.

What Autechre actually said

Be careful what you claim for Sean Booth and Rob Brown. Their catalogue runs from tight, electro-minded programming to long abstract pieces, and no one method covers it. In Sound On Sound's 2004 interview, Booth says beats that seem to move around in time and space are "not random": they follow "sets of rules" the duo have "a good handle on". The same interview has him saying the album Draft 7:30 was mostly written note by note, with one track, "Reniform Puls", using generative material from Max. Rules in one place, hand placement in another, and nothing there says chance never comes into it.[1]

Sound leads too. In a Future Music interview from around 1994, republished by MusicRadar in 2021, Brown says a handful of sounds will "dictate what kind of rhythm we use". And in a 2010 Tiny Mix Tapes interview he describes turning an idea into numbers in Max, with an algorithm then adding "a bit more flair" or deviation.[2][3]

Our reading, which is not theirs: the rule comes after a musical decision, and a person still decides what survives. The skill is designing a system and then listening to it, not handing authorship to the machine.

A five-step click against a 16-step bar

What follows is invented for this article. It is not an Autechre pattern or a guess at one, and it needs only a drum sampler and a MIDI clip. Count sixteenth steps 1 to 16 in each 4/4 bar at any tempo you like.

  • Anchor, never changing: kick on steps 1 and 9, clap or snare on steps 5 and 13.
  • The rule: a short, high, tight click that plays every five sixteenths. Make a one-note clip with a loop length of five steps and let it run across the bars.

Five doesn't divide 16, so the click and the bar can't agree until they have each completed a whole number of cycles. That first happens after 80 steps, the lowest common multiple of 5 and 16: five bars, or sixteen clicks. Where the clicks land, by bar:

  • Bar 1: steps 1, 6, 11, 16.
  • Bar 2: steps 5, 10, 15.
  • Bar 3: steps 4, 9, 14.
  • Bar 4: steps 3, 8, 13.
  • Bar 5: steps 2, 7, 12. Bar 6 starts the cycle again.

If your DAW makes a clip that isn't a bar long awkward, skip the loop length: draw these click positions across five ordinary 4/4 bars and loop that block instead. The maths is identical.

After bar 1, every click lands one step earlier than its counterpart in the previous bar. Bars 1 to 4 each have one click on a kick or clap (steps 1, 5, 9 and 13 in turn), and bar 5 has none, so that is the bar to listen to for the click on its own. Three kinds of thing are going on at once: the anchor never moves, the click always sits exactly five steps from the last one, and the relationship between them changes every bar. That is the whole trick, and every part of it is something you chose.

Change one thing: the accent

Now lengthen the clip loop to 15 steps, with clicks on steps 1, 6 and 11 of the loop, and make only the first one loud. Put it near velocity 110 and the other two near 70. The clicks haven't moved, so what stays stable is the five-step spacing. What changes is which click is stressed: every third one. Fifteen steps against a 16-step bar means the accent falls one step earlier each bar, from bar 1 step 1, to step 16, to bar 2 step 15, and so on. It returns to step 1 after 240 steps, which is 15 bars, the lowest common multiple of 15 and 16.

Then let the sound do some work. Route velocity to the click's filter cutoff or pitch, so the accent is brighter or higher. The grid hasn't changed, but the ear now follows the bright click and tends to hear a three-note figure that drifts against the kick, which is a different rhythm from the same positions. If it sounds merely fussy, shorten the envelope or drop the pitch change; that is how you edit it.

Why this isn't random

As a control, scatter sixteen clicks across the same five bars by hand or with a randomiser, then freeze the result. Same count, same density, and once frozen it is as repeatable as any MIDI clip: you can loop it, edit it and keep it if it works. The difference is structural. In the rule version every click sits five steps from the last, so the relationship between click and bar is something you can state and then change. Lengthen the loop and the accent slips a step each bar. A frozen scatter has no such relationship to adjust; you can only move clicks one at a time, and nothing in it tells you which move is the right one.

None of this makes chance forbidden. A little probability on one parameter, a hat velocity, say, can be musical when it sits on top of structure you chose. The mistake is hoping Randomise will compose for you. Nor is complexity a virtue in itself; five bars of a rule that isn't doing anything for the track can be cut.

For the loop problem this feeds into, see why an eight-bar loop isn't a track. A cycle that takes five bars to come round is one way to make the loop itself keep moving.

Make a rule, hear five bars

Make one rule. Hear five bars. Change one parameter: the accent, the click's pitch, or a mute in bar 5. Keep what works and delete what doesn't. That last step is the composing, and no generator does it for you.