>Analysis takes up to ten mins per track, not ten mins of audio content.
>
That's not what the documentation says:

"Before a PUID is available for MusicBrainz or Picard to use, Music Analysis 
must have been performed on a track. MusicAnalysis uses up to 10 minutes of the 
track and examines all sorts of things."

So a closed-source app analyses:
1) the first 2 mins of a track, to generate a fingerprint.
2) the first 10 mins of a track, to generate analysis data.

Something is sent to the MusicDNS server to return a PUID (assume the anaysis 
data created from (2)).  PUIDs are not generated locally, but are generated 
based on the analysis data?

>> Out of the ones that have got PUIDs, there were ~5000 that were listed
>> via running the "List Duplicates" report (generates a playlist).  i.e.
>> different source files in the library that have the same PUID.Are they 
>> different versions of the same track?

Actually, I don't think the "List Duplicates" report is using PUIDs, as it 
seems to be suggesting duplicates on files that don't have PUIDs yet.  Maybe it 
is based on the Analysis data created for each track (or the fingerprint?).

The Duplicates playlist only indicates what tracks it believes may be 
duplicated - it doesn't say what pair of items are considered duplicates.  I 
was ordering by PUID, so I could see songs that shared the same PUID.  These 
generally were the same song, but often slightly different versions, eg. one 
maybe from an album, and a repeat of the same song from a CD single.  Sometimes 
these versions had different lengths, sometimes I know they are different edits 
of the same song (eg. remastered version, remixed, extended, etc).

i.e. a lot of the 2500 songs were legitimate dups, mainly from compilations and 
album+single rips, but some are not legitimate dups.

Phil
_______________________________________________
ripping mailing list
[email protected]
http://lists.slimdevices.com/mailman/listinfo/ripping

Reply via email to