Humanization

Put some samples in a sequence and it can sound stiff, robotic. At least, it can without some intervention. That’s because computers are too on the beat; too tight.

But you can get around that. You can humanize electronic music in lots of different ways and HYPERBLAM lets you combine a number of different techniques. Get loose, get wobbly, get downright shambolic if that’s your bag.

Multiple samples

Drummers are only human, so every hit they make is a little different. Louder, quieter, sharper, duller. If they lift the stick quickly, the sound rings out. If they leave it down, the sound is cut short.

The easiest way to make a sampled hi-hat sound like a real hi-hat is to sample a drummer playing it multiple times. It’s the perfect route to imperfection.

By default, HYPERBLAM instruments will play a different sample from their sample bank each time. The more samples you include, the more variety.

Three buffers converge on an icon of a dice which, in turn, points at a source node.

A tonal instrument like the standard <notes-blam> will select random samples even as it is augmenting their pitch. This multidimensional variation is not common even in expensive, commercial sample-based instruments. At least, not without a lot of configuration. Take a listen to the “pentatonic party” example to see this in action.

Randomized parameters

Including more than a few samples (even when using optimized, next-gen audio formats) is costly in memory. Emulating natural variation is more efficient.

Each sample is represented as an AudioBufferSourceNode, which comes with certain properties called AudioParams. The detune param’ can be adjusted in cents.

A detune parameter, represented as a dice, pointing to a source node.

Applying the same detune value to all samples alike will change the base tuning of the hi-hat, which may be desirable. But fluctuating the detune value is useful for humanization.

By including the eponymous <blam-blam> element, which points to the instrument element’s detune property, and uses value="-25~25", the pitch of each successive hi-hat sound will be detuned by somewhere between -25 and 25 cents.

<blam-blam
  prop="detune"
  value="-25~25"
>
</blam-blam>

This code is taken from the “I wear many hats” example.

The pitch of a hi-hat is affected by how close to the center you strike it. So this is a reasonably realistic but efficient way of introducing humanization. Applying this technique across a bank of alternative samples is doubly effective.

Swing

Consider the following pattern, wherein 1s represent sounds.

1 1 1 1 1 1 1 1

That’s one sound, played 8 times, in a regular—or straight—fashion. Not very inspired. Now consider this pattern, wherein the 0s represent pauses.

1 0 1 1 0 1 1 0 1 1 0 1

Notice how the sounds bunch together into pairs. This gives the rhythm a kind of elasticity sometimes called swing. At faster tempos, it becomes more of a shuffle. In any case, this elasticity gives your music a perculiar kind of sexy and infectious energy.

But we have a problem. In that last example, we had to divide the pattern into sets of three, or triplets. It makes the pattern longer (messing with the overall tempo) and harder to read.

Instead, we can apply swing using the <sequencer-blam> element’s swing prop. This pushes every other beat up against the next. By default, these shifted beats are moved one sixth of one quarter-beat to the right. Effectively, this makes them the third beat in a imaginary triplet spread over two quarter beats (or one half beat).

Diagram comparing straight and swing patterns. In the swing version, a dot representing the second of three beats is swung to the right.

In practice, it turns that original, straight pattern into this:

1  11  11  11  1

Using triplets, either compositionally or algorithmically, may not give you precisely the amount of elasticity you’re looking for.

So, while you can apply swing as a Boolean attribute and get 0.333 of swing, you can adjust the proportion to taste by supplying an explicit value. This is in keeping with classic hardware drum machines and grooveboxes, which typically offer variable swing.

In the “getting ghosted” example, switching on swing applies a swing value of only 0.2. But the change in groove is still noticeable.

LFOs

Low Frequency Oscillators offer humanization opportunities as well. But LFOs can do a lot more besides, so head over to the LFO guide for a proper introduction.