Skip to content

Latest commit

 

History

History
47 lines (32 loc) · 1.7 KB

Readme.md

File metadata and controls

47 lines (32 loc) · 1.7 KB
Raw

FluidSound

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 (https://github.com/MTG/freesound/tree/master/similarity), which uses Essentia and Gaia technologies (http://essentia.upf.edu/).

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 FluidSound.sc 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 (index_database.sh) from command line. It will take some time depending on the size of the database.
  5. Run the server script (run_server.sh)

You should now be able to interact with the database through the same descriptors used in Freesound: https://freesound.org/docs/api/. FluidSound is similar to Freesound.sc (https://github.com/g-roma/Freesound.sc) but focuses on content-based indexing.

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

FLSound.getSound(100, {|result|
    ~snd = result;
    ~snd["name"].postln;
});

~snd.getSimilar(action:{|result|
	result.postln;
	result[1].name.postln;
});


~snd.getAnalysis(action:{|result|
	result.postln;
});


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

FluidSound can be used as a backend for MIRLC: https://github.com/axambo/MIRLC/