Classes & Categories

When a sound is added to a Sononym library, it will automatically be classified & categorized according to our machine-learning model.

NB: This page explains how those categories and classes work. To learn how to sort and filter search results by category, please visit the category filter page.

Classes

A sample can be classified in four different ways:

  • LOOP
  • ONESHOT
  • LOOP + ONESHOT (primarily a loop)
  • ONESHOT + LOOP (primarily a oneshot)

A sound is classified as a LOOP when it seems to have a looping or repetitive nature. Sononym does not (yet) assign categories to Loops.
A sound is classified as a ONESHOT when the sound is non-repetitive. Oneshots are almost always categorized.

Categories

When processing a sample, the software will determine how much it matches the built-in categories. There are dozens of categories, divided into three groups: Tonal, Percussive and Sound Effects (XFX).

Percussive Sounds

This category covers sounds with a mostly ‘percussive’ quality. That is, mostly (or completely) atonal sounds with a clear impact, leading into a sustained decay.

  • Bongos & Congas
  • Claps
  • Cymbal Crashes
  • Cymbal Rides
  • Hats & Shakers
  • Kicks
  • Metal Hits
  • Snares
  • Snips & Snaps
  • Toms
  • Vibraslap & Guiro
  • Vinyl Scratches
  • Wood Hits
  • Zaps & Blips

Tonal Sounds

This category includes sounds with a primarily ’tonal’ character. The sounds can be both simple or complex, sustained or decaying.

  • Bass & LowKeys
  • Blips & HighKeys
  • Leads & MidHiKeys
  • Pads & Textures
  • Stabs & Orch. Hits
  • Triangles & Bells
  • Voice & Acapella

Sound Effects (XFX)

This category covers sound effects of various kinds.

  • Breaks & Smashes
  • Cracks & Rustle
  • Explosions & Shots
  • Nature & Athmospheric
  • Noise & Distortion
  • Sweeps & Lasers
  • Whooshes & Whips

Manually override classes & categories

While the classification system in Sononym can recognize sounds from a wide variety of sources, it certainly isn’t perfect. Therefore, it’s possible to override predictions from the neural network. It’s dead simple - see the animation below:

‘Fixing’ a sample that got misclassified

‘Fixing’ a sample that got misclassified

When you override the value, your choice also affects how similarity search is working. A manually assigned class/category will receive a “perfect” score, and as a result of this, will appear topmost in the search results when you browse by category.

Note that you can only use the pre-defined categories. We will introduce custom tagging at a later point, which will allow you to organize things exactly as you want.


Related pages


Next Chapter: File Formats