05 – cutting Shapes


go forth and listen

It’s not a mad idea to listen to existing forms and try to work out what’s going on with them. I find it useful to engage in purposeful, active tasks and record myself doing them. The tasks themselves have a kind of inevitable form; e.g. starting, working on and finishing. Leaving the office is a favourite of mine.

hunt for form, then sculpt it if you need to

It’s not often that you’ll get a sound recording that is automatically as purposeful and clear as an intentionally shaped composition, but you may be lucky.

If you’re not, don’t worry, there is no shame in editing down a recording into a purposeful and interesting shape, so, go ahead and edit your field recording in order to make a sensible structure that uses 20-30 seconds well (in your opinion).

Label events in the form you have made

Label the form and events stream – this is useful in order to describe to yourself what you think you’re doing and what you feel each sonic event to mean. Here you’re describing to yourself what you infer are the intentions of the sound itself.

identify layers

For this, you switch your listening over from hearing events in time, to listening to layers. Think of this as converting your mode of listening from considering form and structure (i.e. time/rhythm) towards a consideration of timbre and colour and its shifts in-time. How are different layers of sound material competing for attention, allowing one another to coexist? How is the spectra broken up? What’s missing in the spectra?

Label for yourself the function of each layer.

  • what is the function of each layer? How much presence is there?
  • When do things emerge into ‘view’?
  • What else in the soundspace might make that possible to hear?
  • Are things interconnected? See below…

combine your spectral and temporal thinking

  • do different elements of different layers change roles? For example, does the background move to the foreground?
  • When does it happen?
  • Why?

Now what?

In your DAW, it’s wise to setup different conceptual groups of material. There may be different kinds of content depending upon what you’re trying to do. In the end, it’s probably sensible to give yourself some busses to mix, rather than trying to mix what could be over 100 different tracks in the end. You can think of mixing back, middle and foregrounds. Alternatively, you may have something more like a film mix which could require you having mixing tracks for:

  • dialogical/discursive/semantic
  • environmental/contextual/spot-event sounds (diegetic, in-world sounds)
  • atmospheres/feelings/sentiments/drones (non-diegetic, music-like behaviour)


listening to your guide track, there will almost certainly be a sense of space in there. Things aren’t simply in one location all the time, rather they’re moving across the spatial field (as captured by the mics). So, how will you handle the spatial positioning of your different layers and what path is each sound on?


Access the Reaper session I showed and the files I recorded for demonstration purposes in the audiocrafting group folder. It’s called “scenes.zip”.

sonic objects

You have your library, but it probably doesn’t really do much as a series of sounds on their own, so, how will you layer and group these together in order to make individual sonic events?

Making sub/micro libraries from your words

I can’t laud the powers of micro-sound library manipulation enough. As we saw from Krotos’ work, a well organised library of sounds can be really usefully be put to work to make new sounds very quickly. This happens often in game audio too.

Micro-sound editing is extremely laborious and although it’s true to say that we lose detail if we don’t edit precisely (and that perhaps means manually), we can get a very long way with sound file manipulation and reorganisation using batch processing tools to auto-extract chunks of sound and using other tools (such as the one below) to recombine them in interesting ways.

Batch slicing using Aubio

Aubio is a library of very handy tools written in C for a very fast real-time implementation. There is a Python interface to these tools which makes it possible to run aubio on the commandline or shell.

To install aubio on OS X

You should definitely use a prebuilt and compiled version which means you want to get a python package manager installed first, this is called pip. This should be installed automatically with more recent python installs. If you don’t have it, you can follow the instructions here: pip.pypa.io/en/stable/installing/

When your package manager is installed, try typing in the aubio install command:

pip install aubio

or if your installation is python3-based, then

pip3 install aubio


When aubio is there

Try running a command to check it’s happy, type this:

aubio -h

If aubio is comfortably installed, you’ll see the following:

usage: aubio [-h] [-V]  …
optional arguments:
-h, --help show this help message and exit
-V, --version show version
help show help message onset estimate time of onsets (beginning of sound event) pitch estimate fundamental frequency (monophonic) beat estimate location of beats tempo estimate overall tempo in bpm notes estimate midi-like notes (monophonic) mfcc extract Mel-Frequency Cepstrum Coefficients melbands extract energies in Mel-frequency bands quiet extract timestamps of quiet and loud regions cut slice at timestamps
use "aubio --help" for more info about each command

So, try and slice up a sound…

aubio cut fileName.wav -o path/Where/You/Want/Some/Slices/To/Appear

If that worked, then you can run batch on LOADS of sounds all at once:

for f in *.wav; do aubio cut "$f" -o path/to/outouts/; done

Reorganising sounds with a custom tool.

One approach is to invent an event-maker tool and see how you get on. You could explore some of the very powerful tools shown by Krotos last week, or try rolling your own in MaxMSP. An example application is uploaded here: www.dropbox.com/s/vfehghs6hle11ql/shapeMaker_OSX.zip?dl=0

Drop a folder of sound files in the top-right area. Ideally these should be very short sounds with a fade at the beginning and end of each file.
  • drag and drop a folder of sounds on the top right
  • set an event duration (how long do you want the gesture to take?
  • how much tail for the event do you need? Will there be some long sounds that will get clipped at the end of the recording if they don’t play to the end
  • hit the clear buttons on the pan and files curves and draw some curves through the library of sound files
  • you can set the file name of the sound you want to record in the little recorder unit at the bottom of the right hand side
  • every time you hit the button at the top left of the shape-maker window (not the init button, the one to its right), a new gesture will be made of the duration that you set in point 2 above
  • specify pan journey through the file
  • use your mic to walk through the sound file collection instead, or play a sound file by clicking the buttons below. To record your sound you’ll need to manually trigger it when using your microphone, but when playing a sound file, it will stop recording at the end of the sound file play routine.
  • have fun…
sound control of the shapeMaker