Classical Music and Roon (very long post)

Hi everyone. After using Roon for a week with a modestly large classical music collection (about 1000 cd’s), I’ve come to both appreciate what Roon is capable of and get incredibly frustrated by some deficiencies.

So first I will list my frustrations and requests for future features.

  1. I do not currently see the ability to edit an album title. This has led to some level of weirdness and frustration. For example, I have a four-volume set of Paul Lewis playing the Beethoven Piano Sonatas on the Hyperion lable. Though I had correctly named these albums (Beethoven Piano Sonatas Vol. 1; Vol.2; etc.) in my iTunes setup, Roon renamed them - Beethoven Piano Sonatas Vol. 1 became “Beethoven 1”; Vol. 2 became “Beethoven 2”; etc. - only Vol. 4 appeared to get the name correct. I hope that the ability to rename an album will appear shortly so that these kinds of oddities can be fixed (or that Roon will update its own metadata to fix them). Similarly, though I had meticulously gone through the track titles to properly identify each sonata’s common number instead of Opus, Roon lost this information for one of the discs of Vol. 2. I know them well enough to select by opus number, but everyone refers to the sonatas by common number unless you are a musicologist!

  2. The inability to rename album titles also means that when you want to just view your list of albums by composers, things get weirdly out of place. I have four discs of Angela Hewitt playing Beethoven sonatas, but three of them are named “Beethoven: Piano Sonatas” while a fourth is named simply “Piano Sonatas”. So when I look at a list of albums by album title, the Hewitt sonata discs are weirdly separated. Yes, I know that I can sort by artist and then they will appear together - except that this brings me to another problem:

  3. Some of the artist metadata is also weird - like listing Ludwig Van Beethoven as one of the “artists” on my Herbert von Karajan two-disc set of Beethoven’s Missa Solemnis and my RCA recording of Heifetz playing the Beethoven Violin Concerto.

  4. Some albums seem to be just totally missing from the Roon database. For example, I have a 11-disc set of Alfred Brendel playing Beethoven’s piano sonatas released by Philips in the 1980’s. Originally, Philips released these as three multi-disc sets, spanning Sonatas 1-12; then 13-25; then 26-32. But I bought them as a single set. Roon’s album database has the 3 individual volumes in it, but not the entire set, and I can see no way to take my entire set (which has discs numbered 1-11 and make them into a single album recognized by Roon, or even to break them apart into the three separate sets that Roon DOES have. Which brings me to a third problem:

  5. When I use the “find album” feature and I find an album with multiple discs, often it lists my tracks out of order. Getting them in order is incredibly tedious, because you can’t simply drag tracks from one spot to another - you have to click on the up/down buttons to move the track. Try moving one of 100 tracks in a multi-disc set from the bottom to the top and see how long this takes. PLEASE implement drag/drop for this feature, OK? The buttons may work fine on single-disc pop or jazz releases, but not on 25-disc sets of Haydn Symphonies!

OK - that lists my current issues with Roon, but I’m still figuring out how to use the software. There is very great potential here, but the classical database functions need some work. I realize classical music indexing is several orders of magnitude harder than for pop/jazz, but there is SO MUCH POTENTIAL here that I beg you to work on this. Roon I think can be the answer all of us classical music collectors/fans have been waiting for, but it’s just not there yet.

But it has so much potential, and it is so good already with pop/jazz that I’ll almost certainly be keeping it and hoping for the best in the future on the classical side.

John Colombo


Thanks for taking the time to write this out. I think you will find this thread to be worth a read. This post, too.

I completely agree with you about the shortcomings.

Some, like (4), are relatively localized user interface warts. Things like this will get attention once our launch craziness calms down.

Some are data problems with our metadata providers, like (2). To some extent, this may be improved by the changes discussed here, since those changes will separate the concept of “album artist” from “main artists associated with an album”.

For (1), you need “Prefer Local Metadata” (a feature that is not released yet, currently being designed) + maybe some basic editing features, too. We are still working out our relationship with metadata editing, and trying to decide how deep of a metadata “grooming” platform Roon should be. One thing is very clear: it’s important to provide a way for people to “get back” any pre-existing data that we may have damaged in the process of metadata enhancement.

(3) is probably the most complex problem of those you mentioned. If our data sources don’t know about the album, it’s difficult for us to know about it. I think many of the metadata providers out there just give up when it comes to classical disc sets.

One idea I’ve been thinking about is attempting some kind of work-level identification when album-level identification fails. Even if we can’t get great album-level data (review, artwork, etc), it would be nice to get the work structure into a reasonable state. That way, if your files started out with reasonable tags + album information, we can at least make the system superimpose work/part structure and composer information on the tracks. This + whatever data is present in your files would be a huge step forward from where you are today.

The most important thing, I think, is that we are working on it. These data modeling problems are really the meat of it for us. You can think of Roon as a sort of research platform for figuring out how to model and navigate music collections–it’s going to grow and change over time as we learn more and do more. It will take time, but we’re not going anywhere :slight_smile:


I have some similar oddities, listed here, and here.

i’m also curious about @brian’s comment here wrt genres being Album or Artist level, rather than track level attributes. does this mean that all tracks on a given Artist/Album are given the same genre?

for classical music, are Styles treated the same as Genres? ie, Chamber Music, Opera, or Choral Music vs Classical, Baroque, Romantic…

Hi Brian.

Thanks for the detailed response. One thing you all might think about at some point is asking your user base if they would be willing to have Roon “upload” their metadata to you. I suspect that with future enhancements like “prefer local data”, you are going to end up with a large group of very sophisticated classical music collectors who will have taken the time to meticulously groom their data. If you could collect all that data from them (give us a free year’s subscription in return or a discount on a lifetime sub), you’d probably end up with one of the best classical music metadata databases in the world. Why not try to take advantage of that in some way?

Meanwhile, just keep the improvements coming. I see so much that is already good and so much future potential that I’m going to stay on board for a while to see what you all come up with.

John C.

i’m also curious about @brian’s comment here wrt genres being Album or Artist level, rather than track level attributes. does this mean that all tracks on a given Artist/Album are given the same genre?

Yes and no. In the data model proper, there is no connection between tracks and genres. However, when you focus on track by genre, we are using the album genre as a proxy, because we don’t have track-level genre data. There are no other situations in the product where tracks and genres interact.

for classical music, are Styles treated the same as Genres? ie, Chamber Music, Opera, or Choral Music vs Classical, Baroque, Romantic…

We do not plan to emphasize genres as a way to browse classical music. They’re not a good fit, and every hierarchy we’ve seen seems to confuse underlying concepts (for instance, Baroque, String Quartets, Symphony shouldn’t ever be the same type of thing–one represents a time period, one is a form, and one is an instrumentation).

For classical browsing, we are representing as much structure as we can in a first-class manner. Information like you mentioned above is organized via three mechanisms separate from Genre: Form, Period, and Instrumentation. These are all work level fields, since it would be inappropriate to call an album “Baroque”. It would also be inappropriate to put them at the track level, since they are really not attributes of a track.

Here are some examples:

A historical note: in most pre-existing software, Genre has ended up as a sort of catch all field for people to build their own classification systems. I think this leads to some sloppy thinking about what a “genre” really is, since it’s often the only classification system available and must be stretched to support these other concepts. One of our goals, especially for classical browsing, is to clean up this mess. That said, we’re working on a solution for bringing user-defined genres from files into the system as a “user defined classification” mechanism that exists side-by-side with our data model.

One thing you all might think about at some point is asking your user base if they would be willing to have Roon “upload” their metadata to you

Absolutely. There are a number of forms this might take, but building a mechanism for getting user-generated content into the system (either automatically, manually, by mining data from metadata corrections/edits, or whatever) is something that we’ve given a lot of though to. Big project, but worth tackling.

I strongly suggest that Roon layout the concept of their classical music plan before any actual implementation so that users can vote on it.

I, like many classical music lovers, spent many hours on making the metadata of classical music as perfect as possible. We need to know your plan before we can modify our existing metadata to make the feature ‘prefer local metadata’ useful. At least for me, I do not buy Roon for perfecting my metadata. I buy it for the extra info coming to each work.

Finally, what do you mean by string quartet and symphony are instrumentation and form? To me, both are form of classical work. They should be under Form for example. Instrumentation should be like piano, strings, winds, orchestra. However, instrumentation is not that important in browsing classical music. I would not put too much emphasis on it. For example, what instrumentation would you put violin sonata or piano wind quintet under?

Period is again very easy. If one knows the composer, one knows the period. You can associate it to composer field. It does not have to be in existing metadata.

what do you mean by string quartet and symphony are instrumentation and form?

“Sonata”, “Symphony”, “Concerto” are forms. They represent information about the internal structure of a work.

“String Quartet”, “Full Orchestra”, “Chamber Ensemble” are instrumentations. They describe the set of musicians that a work was composed for.

A “Symphony” is generally expected to be a work for full orchestra, but not all pieces written for a full orchestra are symphonies–it would be inaccurate to combine the two concepts into one and call “Symphony” an instrumentation.

For example, what instrumentation would you put violin sonata or piano wind quintet under?

We track instrumentation as precisely as possible given the source data. “Piano Wind Quintet” would end up (in the data model) as “1 piano + 5 wind instruments”. Violin sonata would end up as “1 violin + 1 keyboard”. How we choose to render these data modeling concepts into display text is a separate problem from accurately representing the data we have.

Why do you single out string quartets as instrumentations? If you do that, you should do it for string quintet. BTW, some string quintets have two violas and some has double bass. Also one of the Schonberg string quartets have a female voice as well!

It makes me wonder what is your purpose and use of instrumentation. For classical lovers, just looking at the work will know these information. For non classical lovers, they do not give a damn anyway.

@jdcolombo Is the your Alfred Brendel set the Philips box with catalogue number 412 575-2? If so, Roon identified it correctly for me, so I could check the metadata, etc. if you want.

I use Musichi Suite to tag my classical cds. It is the best software I have found for tagging classical. You might want to take a look at it.

1 Like

Musichi indeed does look promising, people on CA are quite fond of it judging by this thread:

Seems like they use their own database, maybe someone to contact for licensing, @danny ? :wink:

Yes, that’s it.

Help appreciated - email me directly at jdcolombo at gmail

John C.

Hi @brian, Not wishing to be argumentative just for the sake of it but isn’t this taxonomy over simplistic? While a string quartet is of course an instrumentation, I think most musicologists would consider that it is also very much a form, at least as much as symphony is. Quoting from Wikipedia (sorry, lazy). My bolds:

A string quartet is a musical ensemble of four string players – two violin players, a viola player and a cellist – or a piece written to be performed by such a group. The string quartet is one of the most prominent chamber ensembles in classical music, with most major composers, from the mid to late 18th century onwards, writing string quartets.

The string quartet was developed into its current form by the Austrian composer Joseph Haydn, with his works in the 1750s establishing the genre. Ever since Haydn’s day the string quartet has been considered a prestigious form and represents one of the true tests of the composer’s art. With four parts to play with, a composer working in anything like the classical key system has enough lines to fashion a full argument, but none to spare for padding. The closely related characters of the four instruments, moreover, while they cover in combination an ample compass of pitch, do not lend themselves to indulgence in purely colouristic effects. Thus, where the composer of symphonies commands the means for textural enrichment beyond the call of his harmonic discourse, and where the concerto medium offers the further resource of personal characterization and drama in the individual-pitted-against-the-mass vein, the writer of string quartets must perforce concentrate on the bare bones of musical logic. Thus, in many ways the string quartet is pre-eminently the dialectical form of instrumental music, the one most naturally suited to the activity of logical disputation and philosophical enquiry.

Indeed instrumentation and form are very often inextricably inter-dependent so I think any attempt to treat them as completely orthogonal concepts, as your post seems to imply, is fundamentally flawed.

True, except when one wants to browse, say, 20th C Chamber music. In this case, seeking a particular composer is not helpful. I like the idea of period/form, eg Baroque/Chamber or Romantic/Symphony.

Why not simply ID as “Chamber Music”?

The more I use Roon with classical music, the more I see the need for an alternative to AMG (Rovi).

For many classical box sets, AMG/Rovi has no information.

For example, I am now in the process of ripping Lorin Maazel: The Complete Early Recordings on Deutsche Grammophon, and am finding that AMG does not identify almost all of the disks.

I have had similar problems with other sets from Universal Classics, Sony Classical and Warner Classics. I had the same problem ripping the Markevitch Icon set on Warner Classics.

I frequently get many more hits using Gracenote, but I know that they don’t have as much rich information as AMG. However, if Rovi can’t identify the album at all, there is no rich information.

Are there other alternatives that Roon can use?

I’ve just had a similar problem with the new Stravinsky Complete Columbia Album Collection - no metadata at all on Allmusic, whereas pretty good track and artist metadata on Gracenote. A more classical friendly alternative would be really welcome.

How about ripping in dBPoweramp which supports many metadata sources, and then letting Roon rely on the info in the files? Does that work for you?

(I don’t do much classical myself, but I do recognize your problem.)

dBPoweramp made quite a big thing about including the metadata from SonataDB in their suite of providers