Surfing Random Streams

Back in episode 4 of this tutorial series we took a brief look at randomisation whilst coding up some sizzling synth riffs. Given that randomisation is such an important part of my live coding DJ sets I thought it would be useful to cover the fundamentals in much greater detail. So, get your lucky hat on and let’s surf some random streams!

There is no random

The first thing to learn which might really surprise you when playing with Sonic Pi’s randomisation functions is that they’re not actually really random. What does this actually mean? Well, let’s try a couple of tests. First, imagine a number in your head between 0 and 1. Keep it there and don’t tell me. Now let me guess… was it 0.321567? No? Bah, I’m clearly no good at this. Let me have another go, but let’s ask Sonic Pi to choose a number this time. Fire up Sonic Pi v2.7+ and ask it for a random number but again don’t tell me:

print rand

Now for the reveal… was it 0.75006103515625? Yes! Ha, I can see you’re a little sceptical. Perhaps it was just a lucky guess. Let’s try again. Press the Run button again and see what we get… What? 0.75006103515625 again? This clearly can’t be random! You’re right, it’s not.

What’s going on here? The fancy computer science word here is determinism. This just means that nothing is by chance and everything is destined to be. Your version of Sonic Pi is destined to always return 0.75006103515625 in the program above. This might sound pretty useless, but let me assure you that it’s one of the most powerful parts of Sonic Pi. If you stick at it you’ll learn how to rely on the deterministic nature of Sonic Pi’s randomisation as a fundamental building block for your compositions and live coded DJ sets.

A Random Melody

When Sonic Pi boots it actually loads into memory a sequence of 441,000 pre-generated random values. When you call a random function such as rand or rrand, this random stream is used to generate your result. Each call to a random function consumes a value from this stream. Therefore the 10th call to a random function will use the 10th value from the stream. Also, every time you press the Run button, the stream is reset for that run. This is why I could predict the result to rand and why the ‘random’ melody was the same every time. Everybody’s version of Sonic Pi uses the exact same random stream which is very important when we start sharing our pieces with each other.

Let’s use this knowledge to generate a repeatable random melody:

8.times do
 play rrand_i(50, 95)
 sleep 0.125
end

Type this into a spare buffer and hit Run. You’ll hear a melody consisting of ‘random’ notes between 50 and 95. When it’s finished, hit Run again to hear exactly the same melody again.

Handy Randomisation Functions

Sonic Pi comes with a number of useful functions for working with the random stream. Here’s a list of some of the most useful:

Check out their documentation in the Help system for detailed information and examples.

Resetting the Stream

Whilst the ability to repeat a sequence of chosen notes is essential to allow you to replay a riff on the dance floor, it might not be exactly the riff you want. Wouldn’t it be great if we could try a number of different riffs and choose the one we liked best? This is where the real magic starts.

We can manually set the stream with the fn use_random_seed. In Computer Science, a random seed is the starting point from which a new stream of random values can sprout out and blossom. Let’s try it:

use_random_seed 0
3.times do
  play rrand_i(50, 95)
  sleep 0.125
end

Great, we get the first three notes of our random melody above: 84, 83 and 71. However, we can now change the seed to something else. How about this:

use_random_seed 1
3.times do
  play rrand_i(50, 95)
  sleep 0.125
end

Interesting, we get 83, 71 and 61 . You might notice that the first two numbers here are the same as the last two numbers before - this isn’t a coincidence.

Remember that the random stream is just a giant list of ‘pre-rolled’ values. Using a random seed simply jumps us to a point in that list. Another way of thinking about it is to imagine a huge deck of pre-shuffled cards. Using a random seed is cutting the deck at a particular point. The fabulous part of this is that it’s precisely this ability to jump around the random stream which gives us huge power when making music.

Let’s revisit our random melody of 8 notes with this new stream resetting power, but let’s also throw in a live loop so we can experiment live whilst it’s playing:

live_loop :random_riff do    
  use_random_seed 0
  8.times do
    play rrand_i(50, 95), release: 0.1
    sleep 0.125
  end
end

Now, whilst it’s still playing, change the seed value from 0 to something else. Try 100, what about 999. Try your own values, experiment and play around - see which seed generates the riff you like best.

Bringing it all together

This month’s tutorial has been quite a technical dive into the workings of Sonic Pi’s randomisation functionality. Hopefully it has given you some insight into how it works and how you can start using randomisation in a reliable way to create repeatable patterns within your music. It’s important to stress that you can use repeatable randomisation anywhere you want. For example, you can randomise the amplitude of notes, the timing of the rhythm, the amount of reverb, the current synth, the mix of an FX, etc. etc. In the future we’ll take a close look at some of these applications, but for now let me leave you with a short example.

Type the following into a spare buffer, hit Run, and then start changing the seeds around, hit Run again (whilst it’s still playing) and explore the different sounds, rhythms and melodies you can make. When you find a nice one, remember the seed number so you can get back to it. Finally, when you’ve found a few seeds you like, put on a live coded performance for your friends by simply switching between your favourite seeds to create a full piece.

live_loop :random_riff do
  use_random_seed 10300
  use_synth :prophet
  s = [0.125, 0.25, 0.5].choose
  8.times do
    r = [0.125, 0.25, 1, 2].choose
    n = (scale :e3, :minor).choose
    co = rrand(30, 100)
    play n, release: r, cutoff: co
    sleep s
  end
end
live_loop :drums do
  use_random_seed 2001
  16.times do
    r = rrand(0.5, 10)
    sample :drum_bass_hard, rate: r, amp: rand
    sleep 0.125
  end
end