rediscover  - the smart cover version discovery service
---------------------------

For now available as desktop version (for mac OS X)
working on android version...
interested to use it or support with dev: write me to joro.dzhambazov@gmail.com

dependencies:
-pyaudio (to bridge to PortAudio)
-pyspotify
-libspotify

steps
I. find covers

1)
Cover versions for a song title are queried from a file: the MSD-SecondHand
http://labrosa.ee.columbia.edu/millionsong/secondhand
they are around 5000 songs with 3-4 cover versions on average 

return: list of MSD (echonest) track IDs

2) match form MSD to musixMatch (MXM) IDs
return: list of MXM song IDs

3) discover most representative audio segments (thumnbnails) and get their timestamps 
(uses with musixMatch API to get snippet from lyrics and then gets timestamps from available subtitles) 
sort by duration of thumbnails

return:   list of sorted songIDs and corresponding ts-s  

II. play with pyspotify: 
1) add spotify IDs from musixmatch

2) play: 
cannot play form given ts to other ts, so we use player.seek(), timer.wait() and player.unload()


Spotify API requires some tweaking: 

playing music:
https://pyspotify.mopidy.com/en/latest/quickstart/#playing-music

redisCover uses PortAudio 
https://pyspotify.mopidy.com/en/v1.11/audiosink/
although it has some issues:
jodal/pyspotify#132  


------ 
workplan: 

port to android
https://developer.spotify.com/technologies/spotify-android-sdk/tutorial/
and
https://developer.spotify.com/android-sdk-docs/com/spotify/sdk/android/playback/Player.html


thumbnails from musixMatch do not have smae lyrics: 
ideas:
 cluster thumbnails: exclude songs with  thumbnails outside of cluster
 as similarity metris use levenshtein distance 
 
   
optimize data structures in beginning

see as well issues