This compilation box set has 67 discs, and was issued in 2014. I’ve so far ripped 20 of the discs. I manually added/corrected metadata as best as I could. Roon has two plausible candidates for identifying the album, but when I try to, the tracks get all jumbled in a very bad way, even though they are numbered properly in the metadata (track and disc number).
I have this set. I’ve not found any metadata for it. I’ve entered it all manually.
I’ll see if I can get an identification. Thanks for the heads up.
Discs shouldn’t get jumbled if you have CD1 in folder name and disc 1 of 67 in the tags.
I wish… I renamed the folders appropriately, but the tracks are still utterly jumbled. In addition, Roon decides that there are two albums, one with 12 discs and the other with 4 (even though there are actually 20 discs). I wish I could force Roon to just believe the metadata on the files, rather than trying to figure out a structure and failing utterly.
Managed to fix some of the problems, but by no means all. Roon still breaks the collection into two pieces, the smaller of which is still badly jumbled. Is there a reliable way of adding metadata that forces Roon to group tracks into works?
OK, I managed to merge and renumber the 20 discs I have ripped so that the tracks are all in order. But many of the track titles are from somewhere else than the metadata, even though the collection is officially unidentified. Most works are not grouped, and composers are missing (even though the track metadata is correct).
Cleaned out embedded metadata further, added WORK tags, rescanned, re-identified. Still fails to group most of the works. Any thoughts on how I might get Roon to group the works?
Are you removing then cleaning the database then reimporting? If you don’t then your old identification and probably grouping of works will not be changed.
I specifically said folder and CD1 of 67 in tags because in my experience for big boxes with metadata tags, it’s worth doing belt and braces. But that won’t make any difference if you don’t remove (change folder name to “temp”), clean out database, then reimport (Ctrl-Z folder name).
My copy of this box is all fine. It is possible!
Did what you recommended. I have all the tracks in the right order. Discs 1-4 have work groupings after I associated the album to the most likely identification, but the tracks on the other discs are not grouped by work, even though I have WORK metadata on all the tracks. Is there any metadata trick that will achieve work grouping?
My understanding is that if you have WORK you have to have PART as well for it to work. Do you have PART as well?
I also have learnt (the hard way) that if you use WORK/PART on one work on a disc, Roon fails to group all the other works on the disc unless you also use WORK/PART for those. In other words you have to use it on the whole disc or not at all. (I use the word “disc” advisedly, in that I have only noted this on single-disc albums. I don’t know if the same thing will be the case across an album which is multi-disc.)
Maybe the Roon team will advise if my understanding is correct.
Thank you again, that finally starts making sense now that I discovered http://kb.roonlabs.com/File_Tag_Best_Practice, which is not so easy to find… Adding PART tags now, still a long way to go…
Thanks again! Making a lot of progress in getting the metadata clean for Roon. It’s quite a pain to have to go through the whole dance of moving the boxset folder away from Roon storage, rescan storage, clean it, move the folder back to its normal place, and rescan storage to fix the effects of incomplete/bad metadata or files that failed to be ingested (who would have known that “?” is forbidden in Roon filename path?). I’m learning a lot, but it would have been easier if 1) all the documentation on metadata preparation was together and linked to clearly from Roon documentation; and 2) totally rebuilding a boxset’s Roon data did not require the dance above (maybe it does not, but I’ve not been able to succeed with anything else).
Just to close this off, thanks to @Ludwig for all his advice and to the authors of the “File Tag Best Practice” page, I’ve got a good process to ingest this boxset and partition it into compositions. It would be great if there was “official” metadata for the boxset, but it’s obviously a lot of work since Sony didn’t bother to supply it. Once I have finished ripping the whole set I’ll have a not horrible set of metadata, but unfortunately there’s too much variation in work and part names for me to hope that “magically” I’d get a merger of all the versions of several compositions that appear in this set and also in several other albums I own. Oh well.
I also have learnt (the hard way) that if you use WORK/PART on one work on a disc, Roon fails to group all the other works on the disc unless you also use WORK/PART for those. In other words you have to use it on the whole disc or not at all. (I use the word “disc” advisedly, in that I have only noted this on single-disc albums. I don’t know if the same thing will be the case across an album which is multi-disc.)
Ah - that could explain some of the weird behaviour I am noticing from time to time. I try to have WORK/PART always populated. Only problem is with opera, where I am too lazy… But I’ll need to change that since I will end up with 250 recordings of Rigoletto because every single track is then reckognized as the work 
I’m a bit unsure about this. What if the disc has both multipart WORKs and single-part WORKs, for which there is no PART value that makes sense? I’ve been leaving out the PART for single-part WORKs with no apparent ill-effects so far.
Clearly you won’t have problems with grouping of single track works.
With a disc with two multipart works on, using WORK/PART on one only causes the other to fail to group, even if it had previously correctly grouped. I can’t say more than that.
Please try it and let us know what you find. Maybe this “feature” is more complex in its behaviour.
This is not good. You should be able to achieve this. And without any Work/Part tags. I’ll put up some screen shots when I can get to my laptop.
I have that merger on every of my 6500 albums. It’s crucial to making Roon work for classical.
Not even close. For example, I have multiple recordings of several Bartók compositions, and the ones from the boxset, with WORK and PART names pretty close to the text in the boxset booklet, did not merge with the existing ones. I had to merge them by hand with composition merge (which seems pretty handy). Maybe I’d have to carefully make sure the WORK and PART metadata for the new recording of a WORK is exactly the same as in the existing recordings, but that would make ripping unidentified recordings even more painful than it is already (oh for a ripper that would use metadata from Roon automatically…)
Update: Not even for identified albums! For example, I have three recordings of Boulez’s “Anthemes II” (or “Anthemes 2”), two of those recordings in different DGG albums, and they aren’t all merged. I’m not totally surprised by this, I’m pretty familiar with the state of the art in algorithms for “record reconciliation” and music is a pretty challenging domain for doing it very accurately.
I’m holding a sleeping baby. I’ll get back to you asap… (I have all those Anthemes Ii recordings as well…)
Enjoy, they grow up so fast 
I’m dealing with a bit of insomnia from allergy drugs (it’s a crazy blooming Spring around here…) and I need to try to go back to bed.
A quick guide to achieving this. It’s important to first understand that Roon has an idea of a “canonical” title of a work. This is the title you will see when you look at a Work Details Page which has a description of the work. It comes from Tivo/Rovi (viewable on Allmusic)
For identified albums:
-
Make sure you have the best possible identification, preferably a Rovi one. If you have a non-Rovi identification you will not get the canonical title, then you will not get an automatic match. Telling which are Rovi identifications is a dark art, but one clue is that the composer will NOT be the Album Artist.
-
If only non-Rovi identification exists, you will get a non-canonical work name. This will probably appear in the composer work list separately from any Rovi identifications. At this point you can sort the composer work list by name (or opus), and find the two works then MERGE them. (Highlight both, Merge button appears top right.) Then all will be well.
For unidentified albums:
-
(The old way) Make sure you always use precisely the Work Titles from Allmusic.com. (These are the Roon canonical names, at least for all major works.) NOT necessarily those from the booklet as you mentioned above. If you use precisely the same Work Title, with the correct composer credit, the works will match up with other performances in your library. (Be careful of spaces and cases.)
-
(The new way) Use any title but make sure the opus number is in there and correct. Roon now has an automated system which related works by opus number. It means with quite basic metadata you can get full matching automatically in Roon. It was a miracle when this feature was introduced. (It has problems if there are typos in the metadata sources, such as Bartók Concerto for Orchestra/Piano Concerto no. 3 which is confuses because of a typo somewhere.)
-
(The lazy way) Use any title you want, and just merge it with the canonical one in the Composer Details work list.
One other thing: About the work descriptions you seen in Roon. If you don’t have an identified album in your library with the same work in it, you will not get this work description, even if you use the exact same work title string as in Allmusic. (It has been proposed that Roon should pull down this info for unidentified albums, but it is not yet there.) My work around (when I care enough) is to find a Tidal album with the work, add it, identify fully, maybe also merge works if necessary, then delete the Tidal album again. The association between the work (from an unidentified album) and the description text, remains after deleting the Tidal album. Kludgy but it works.
Disclaimer: Roon’s metadata system is evolving all the time. Things change. And nothing is quite what it seems. Metadata doesn’t necessarily entirely come from one source even if it looks like it does…
HTH
1 Like