SAWA-Recommender is a content based music recommendation service.
It uses audio analysis (timbre feature extraction) to find songs in our database
which sound similar, in some sense, to one or more uploaded music files.
Most notably, it finds songs which have similar instrumentation or similarly
sounding dominant instruments such as lead vocals, saxophones, violins etc...
itself is a playlist generator which uses the same analysis as this service. It is available as a cross platform plugin for the iTunes
SAWA-Recommender uses a database of timbre features extracted from about 150.000 tracks in a distributed way, by users of SoundBite.
This database mainly consists of popular music, however the similarity algorithm itself proved valuable also in the context of classical recordings.
SAWA-Recommender requires an audio example to be used as a seed for similarity search.
You can upload one or more audio files, however at least one file is required for the service to operate.
A wide range of audio file types are recognised, including aiff, m4a, mp3, mp4, ogg, voc, w64, wav
To use the service, upload a file and once completed, press the find similar songs
button displayed next to the file name.
The uploaded files are identified using the MusicBrainz
database. Where available, the results of identification can be seen by clicking the metadata icon next to the file name.
In case of more than one uploaded seed track, SAWA-Recommender can be configured to search the database for songs that are similar to all uploaded files, and combine the search results,
or it can combine the features of the seed tracks, and execute a single query using the combined seed. These options are available under Advanced search options