Genpuid advice

[Philip Meyer]

As I recently backed up my music library, I thought I'd have a go at using genpuid to archive fingerprint/puid data.

I thought I'd run it on a single folder to start with as a test, using the following:

genpuid.exe 57aae6071e74345f69143baa210bda87 -m3lib="C:\Documents and Settings\Phil\Application Data\MusicIP\MusicIP Mixer\default.m3lib" -archive -r "M:\Music\Phil's Music\Ambient\Alio Die & Saffron Wood" -logex

This reported:
Songs to be scanned: 3

Unfortunately this seems to have kicked off the archiving process on ALL files (assume the m3lib param therefore overrides the folder I specified).

Anyway, it's been processing my files for some time, and I haven't noticed any issues so far.

I was worried in that I heard that genpuid could cause corruption of some mp3 files with large id3 tag blocks. Any ideas how I can confirm that nothing is corrupted? I can restore any corrupted files from a backup, but how would I know (unless I compare every tag before/after archive analysis, and play every song!)?

Also, what tags would this have written to the files? I notice several tags in mp3 files:

MUSICIP DATA
MUSICMAGIC DATA
MUSICMAGIC FINGERPRINT

For flac files I have:
ANALYSIS
FINGERPRINT
MUSICIP_DATA
MUSICIP_PUID

All values are different.

I had previous archived some analysis via the MusicIP GUI, perhaps Genpuid and MusicIP write different tags? Are some redundant?

Phil

[audiomuze]

The -r parameter tells it to recursively processes all subdirs.

[audiomuze]

I was worried in that I heard that genpuid could cause corruption of some mp3 files with large id3 tag blocks. Any ideas how I can confirm that nothing is corrupted? I can restore any corrupted files from a backup, but how would I know (unless I compare every tag before/after archive analysis, and play every song!)?

From memory the problem was that it would truncate tags that had multiple values e.g. genre=Rock\\Acoustic Rock would be rewritten as genre=Rock. I believe this may have been an issue with genpuid < 1.3. I've run genpuid 1.4 (Linux) against mp3 files that include extensive tags and never experienced an issue.

No idea though how you'd check the metadata other than manually loading some files and checking the tags against the backups. If I were you I'd hit Ctrl-C to terminate the analysis and experiment on two or three mp3 albums copied somewhere for this purpose.

FWIW I've just tried genpuid 1.4 (Linux) on an mp3 album including multiple values for a tag. Analysis tags were written and existing tags weren't truncated.

[audiomuze]

Also, what tags would this have written to the files? I notice several tags in mp3 files:

MUSICIP DATA
MUSICMAGIC DATA
MUSICMAGIC FINGERPRINT

For flac files I have:
ANALYSIS
FINGERPRINT
MUSICIP_DATA
MUSICIP_PUID

All values are different.

I had previous archived some analysis via the MusicIP GUI, perhaps Genpuid and MusicIP write different tags? Are some redundant?

No, different tags are used for mp3 vs FLAC files.

[Philip Meyer]

>From memory the problem was that it would truncate tags that had
>multiple values e.g. genre=Rock\\Acoustic Rock would be rewritten as
>genre=Rock.
>
That's what I experienced when I tried it in the past (several years back). But when I recently tried to archive analysis within the MusicIP GUI, I tried an mp3 file with multiple genres and artists and there wasn't a problem (maybe since I switched to id3v2.4?).

GenPuid stopped on its own with a message "Terminated" - it didn't give a reason/error mesage. Roughly half of the music library was processed.

I looked in the MusicIP GUI, and ordered by PUID. I see some duplicate PUID's for songs with the same title, but are actually different versions (eg. one version on a compilation and one of the original album, but with different lengths). PUIDs have always worried me, in that they may be used to re-associate stats (if file path changes), but if PUIDs are not unique, the stats will be applied to the wrong file.

I might remove the PUID tags.

So next question, how do I archive analysis without storing PUIDs in the future?

[Philip Meyer]

>> I had previous archived some analysis via the MusicIP GUI, perhaps
>> Genpuid and MusicIP write different tags? Are some redundant?
>No, different tags are used for mp3 vs FLAC files.

But mp3 files now have MUSICIP DATA and MUSICMAGIC DATA. They have different values. I guess one was stored with MusicIP GUI, and one was stored with Genpuid (not sure what tags each tool generated).

Not sure which one would be re-used if I had to re-import into MusicIP (probably depends on whether I were to use MusiciP headless webUI, MusicIP GUI, or Genpuid), or if there is a difference in the quality of the analysis between the two.

Similarly for FLAC, I have ANALYSIS, FINGERPRINT and MUSICIP_DATA, and MUSICIP_PUID.

[audiomuze]

PUIDs have always worried me, in that they may be used to re-associate stats (if file path changes), but if PUIDs are not unique, the stats will be applied to the wrong file.

I might remove the PUID tags.

So next question, how do I archive analysis without storing PUIDs in the future?

don't think puid's are used as unique identifiers, there's a separate musicbrainz tag for that. From memory there's a parameter you can pass genpuid to inhibit writing of puid's. Can look in morning for you

[Philip Meyer]

>don't think puid's are used as unique identifiers, there's a separate
>musicbrainz tag for that. From memory there's a parameter you can pass
>genpuid to inhibit writing of puid's. Can look in morning for you
>
I thought PUID stood for Portable Unique IDentifier.

The TrackStat wiki page states: "If you have MusicBrainz tags in your music it will even survive a situation where you have moved or renamed your music files, without MusicBrainz tag only statistics for unmoved files with the same name as before will survive."

I don't know what actual tag this looks for - I think MusicBrainz uses PUIDs?

PUIDs are generated I think dependent on roughly the first 8 mins of music. Fingerprint analysis is from 2 mins of music. Neither seem to be guaranteed to be unique, although PUID's seem to suggest they are unique.

I seem to have ~5000 tracks that do not have unique PUIDs.

Phil

[Philip Meyer]

I've determined that Archive Analysis within MusicIP GUI stores MUSICMAGIC DATA and MUSICMAGIC FINGERPRINT within mp3 files. Not sure what the purpose/difference is between those two tags. Another problem seems to be where archive analysis has already occurred - I can't get PUIDs to be updated into the tags.

GenPuid therefore must instead be writing the MUSICIP DATA tag - it doesn't detect when MUSICMAGIC DATA or FINGERPRINT tags already exist, and adds other tags (different tag names containing different values).

Therefore I don't know what benefit there is using genpuid to create MUSICIP DATA or PUID tags (if MusicIP uses MUSICMAGIC DATA tag instead).

I'm confused!

Phil

[erland]

I thought PUID stood for Portable Unique IDentifier.

The TrackStat wiki page states: "If you have MusicBrainz tags in your music it will even survive a situation where you have moved or renamed your music files, without MusicBrainz tag only statistics for unmoved files with the same name as before will survive."

I don't know what actual tag this looks for - I think MusicBrainz uses PUIDs?

TrackStat uses Musicbrainz Id tags, not PUID's. PUID's are just a way for the Musicbrainz tagging software, for example picard, to find the appropriate Musicbrainz Id more automatically. You need the Musicbrainz tag to get the extra functionality in TrackStat and I think neither MusicIP nor genpuid will write any Musicbrainz tags. To get Musicbrainz id's you need to use puddletag, picard or some other tool that support retrieval of Musicbrainz tags.

PUIDs are generated I think dependent on roughly the first 8 mins of music. Fingerprint analysis is from 2 mins of music. Neither seem to be guaranteed to be unique, although PUID's seem to suggest they are unique.

I believe PUID's are almost but not completely unique. Musicbrainz switched from TRM to PUID in their picard tagger because PUID's were more unique.

I seem to have ~5000 tracks that do not have unique PUIDs.

Sounds like a lot, I've got a feeling something is wrong if this really is different songs.

Just out of interest, are they very similar or are we talking about completely different tracks with a lot of silence in the beginning or some other kind of similarity ?

Are we talking about cuesheets ?