How often do you see the terms ‘humanising’ and ‘randomising’ being used interchangeably? Or maybe you’ve seen someone ask how to make something sound more natural or human, and someone else suggests adding random variations to timing and/or velocity? Perhaps you’ve tried adding random variations yourself, only to end up with something that doesn’t sound any more ‘human’ – just sloppy.
Of course, variations to timing and timbre are key to a natural human performance. The lack of variation is one of the very defining characteristics of samples and drum machines (when compared to human performance).
The important difference, however, is not merely that a human performance has variations – but that these variations have a pattern. They’re not completely random.
If you analyse a human drum performance, for example, you’ll find that some notes are always louder than others, some notes are always earlier than others. Other notes are always quieter, others are always later. This is groove. It’s the complex combination of note emphasis, push and pull of timing, and rhythmic mode (which beats in the bar are usually sounding, and which beats are usually not sounding).
Of course, there will also be random variations as well – a human is not a machine! The variations will not only be in timing and velocity, however. There will be tempo variations. There may even be pattern variations (where some bars have a slightly different note pattern). These random variations, however, are not the defining characteristic of a human performance, and are often not the desirable characteristic we are looking for when we want to create the illusion of a human performance.
If not random, then what?
So, if we’re going to add some deliberate variations to timing and velocity, what are we going to do? This is very much a choice that depends on personal taste and the needs of the song. Generally, I find that the following work for me:
- Stronger velocity on the beat (and weaker velocity between the beats). This works when a part needs to be stable and predictable, supporting the song.
- Stronger velocity on off-beats (between the beats) for foreground parts or transitions (such as drum fills).
- No timing offset on the first beat. This just makes it easier to think about where the bar starts.
- No timing offset on the beat (1, 2, 3, 4). This provides a solid, stable beat.
- Notes between the beats to be slightly late. You might already recognise this as swing or shuffle, but I’m talking about doing this much more subtly.
My usual groove within a beat is [0,20,10,20] – assuming 120 ppq. This means that instead of every 16th note (semiquaver) being 30 ticks, the first two semiquavers are 40 ticks long, and the next two are 20 ticks long. This means that there’s a bit of lag between the beats. The effect of this is that:
- Strong notes on the beat have rock solid timing.
- Strong notes on the beat have a bit of extra space after them, making them sound a bit bigger.
- Weak notes lead up to a strong note on the beat sound like they ‘speed up’ leading up to the strong note, adding excitement and emphasis.
- Strong notes off the beat have a bit more funk and groove.
Combined with lower velocities for the offset notes, this is usually not audible as an obvious swing or shuffle. For me, it adds just enough groove that I often don’t feel the need to add additional variations – including random variations – to make a part sound human.
Humanising is more than just adding random variations to timing and velocity (volume and tone). The changes must be musical.
 PPQ: Pulses Per Quarter. One beat (crotchet) is 120 ‘ticks’. Half a beat (quaver) is 60 ‘ticks’. A bar (semibreve) is 480 ‘ticks’.