No description, website, or topics provided.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


FluidSound is a research prototype that allows you to index large databases of sounds and interact with them via content-based descriptors from the SuperCollider language.

It is an adaptation for local databases of the Freesound similarity server (, which uses Essentia and Gaia technologies (

How to use

The server can be executed using a docker image for simpler handling of dependencies. The following steps have been tested on an OSX machine:

  1. Install docker
  2. Install the class in the SuperCollider extensions folder. Recompile the class library.
  3. Copy some audio files to the db/audio folder. Make sure the index folder is empty.
  4. Run the indexing script ( from command line. It will take some time depending on the size of the database.
  5. Run the server script (

You should now be able to interact with the database through the same descriptors used in Freesound: FluidSound is similar to ( but focuses on content-based indexing.

Here's a quick test (run each bit separately in supercollider):

FLSound.getSound(100, {|result|
    ~snd = result;



FLSound.contentSearch(target:".lowlevel.spectral_complexity.mean:2.8", action: {|result|
	~snd = result[0];

FluidSound can be used as a backend for MIRLC: