May I confirm: do you/does that mean that - although I dropped my FLAC files into the appropriate directory (macOS 13.4, Roon 2 etc) and watched Roon go from 1/180 to 180/180 in Settings > Library > Background Audio Analysis Speed, Roon did not ‘consult’ AllMusiuc? Or, if it did, it did not find my downloaded product?
[quote]
I said I felt bad in starting this thread. I obviously do not know enough about Tags.
Where and how do I assign such a set of file to such Tags, please, Tony?
Hmmm. This is how they look in Yate before import:
I do not know Yate. I use mp3tag on a PC so I cannot answer all your questions.
Your WORK tag is incorrect. It is the part before the colon in the TITLE tag. So it is simply:
Pelleas und Melisande, Opus 5.
All track items referring to this WORK should be identical. Roon will try and match it with the canonical form on allmusic. So normally I just paste that in to save a lot of bother:
Pelleas und Melisande, symphonic poem, Op. 5
The composition editor is in the main menu on the left. Once there you get to a grid style interface. I don’t know if there is a tutorial or FAQ how to use it. It is a neglected corner of roon that hasn’t been updated in years so you may struggle with it but the basic functionality is straightforward enough.
I don’t know how you would add a Classical tag in Yate.
There used to be regular posters with Classical libraries using Yate who could help you. But many with Classical libraries have drifted away. Maybe someone will turn up.
Forgive me for adding something likely already known to you, but often the referenced metadata varies ever so slightly from purchased downloads (like Presto), so Roon will initially not identify the recording upon adding to the library. I find that I have to edit the album in Roon and manually point Roon to the correct metadata (which in some cases contains no mismatches but Roon was confused anyway). Often there are multiple versions of a recording (CD, digital download) as well, with slight variations in track timings — I wish Roon allowed the user to adjust the tolerance on this.
In my experience, an “identified” album yields far better results than extensive manual editing of metadata, on balance.
One other thing: since the “Classical” genre is important for Roon, if you use custom genres, make sure they are sub genres of “Classical”.
Which is not actually the case, or at least it does not become visible in the UI. I have lots where I added a composer but it’s still not seen as a composition.
In 2022, Jamie got involved in one case but it ended with crickets:
While I suspect it’s more a case of dealing with less-than-perfect metadata is a particularly tough nut to crack. You don’t want to end up coding in reams of exceptions to your automated rules for dealing with ingesting the metadata.
Yes, it is quite an art form judging the fewest edits needed to get an identification. I took that approach for years but lately, more and more, I find myself tagging local files more completely so that they will work in other players. I started a migration from JRiver more than 5 years a go which I never completed. I sense change in the air with the arrival of consumer AI and want to be ready when the time comes to in turn migrate from roon.
Of course. But that is why consumer AI that can self-learn those exceptions has arrived. That sort of hard coded exception handling approach has reached its limits not just for roon but the entire IT industry. I don’t see any evidence that roon will be able to onboard the AI changes that are coming but I could be wrong. Roon just no longer engages so it is not really possible to say.
In all likelihood, if a business case can be found, such a sea change in music library handling will come from one of the dominant streamers or labels. But you never know. There may already be a couple of enthusiasts beavering away. Stranger things have happened. Even roon started that way. rooExtend has just introduced a new chatGPT based playlist generator and I rather suspect that signals the future. There are clearly teething problems but I don’t think it would be a stretch for others to take that approach with things like search, metadata ingestion, composition identification, box set hierarchical handling and even support for that matter which have long bedevilled roon.
I think that using stochastic parrots to generate a playlist is a completely different kettle of fish from generating 100% accurate metadata from audio tracks. I’m sure that we will see attempts to do this, but accurate results will be a long time coming, and we will see plenty of AI hallucinations in the meantime…
The metadata will never be perfect. Roon should simply make every track that has a composer (even if manually edited) into a composition, and then let me manually merge the composition into an existing one.
Yes, there are certainly well documented teething problems. But these new developments are the beginning of a journey, not the end. Roon used to be like that but it is a mature product now with a different dynamic. Roon’s model that relies on crowdsourced meta-data actually reminds me of the search engine wars when Google arrived on the scene. At that time Yahoo relied on teams of intelligent human experts to curate and categorise web links for its search engine. Google came along and replaced human experts with unashamedly dumb algorithmic web-crawlers. It’s all a long time ago now but we all know how that went.
I sense a similar inflection point now. I rather suspect that this new generation of AI technologies coming out of the labs will be sweeping away business models that cannot adapt as has happened so many times in the past. All technology is trickled out to maximise ROI so it wouldn’t surprise me if hallucination free AI is already to go when the timing is judged right. The unusual speed with which government regulation is stepping in indicates to me that much more advanced AI technologies are indeed ready to go, probably in military applications that government regulators are privy too. I do not believe that in the coming years technical progress will be the bottleneck. Rather it will be societal adaptation.
Anyway, this is clearly way off topic now. The only point I wanted to make was that persistent problems with roon we all experience are persistent precisely because legacy approaches cannot solve them. I have only arrived on this thread because I was cited from 5 years ago on persistent problems with composition identification. Sadly in the intervening period there has been no real progress. It is possible that in the background that roon is ahead of the curve with a new generation of AI tools or maybe other tools but that does seem rather unlikely.
Yes. Its often the quickest fix. After a while you get a feel for it. I don’t know why it happens but I see a few patterns. There are numerous posts. It is not a high priority for roon. One pattern is where there are multiple cataloguing systems. Roon for example will often confuse CD numbers with L numbers with Debussy. You will notice that the allmusic canonical form has both CD and L numbers. Often pasting in the allmusic canonical form is the only way to get roon to correctly identify Debussy compositions. Ravel, I notice, often requires an M number for identification. Other vulnerable composers are Telemann and Haydn where there are very large, very complex catalogues. This is not meant to be exhaustive. Just examples. I like modern compositions and identification is very hit and miss with those to be honest. Personally I am in the habit of replacing most composition titles in my local library as I go along as I really do not want to go through this again with other players. That’s a decision you need to take.
You have 22 recordings from roon’s perspective because it is treating each track as a WORK, not as a PART. There must be something wrong with your work/part edits. The display on the album page should be hierarchical like this. What do you see? I have highlighted where the Classical tag should appear:
Users with large Classical libraries are a very small minority in roon. Some have been very vocal on this forum in the past. Now, not so much as it has become clear that long-standing issues will probably never be addressed.
Have heard back from the developer of Yate and he has a little mini script (an ‘Action’ in Yate) which should do the trick - I think!
Which I achieved by Editing the Album (… > Add to tag).
It would be good to be able to do this before Import (in Yate), which prompts me to ask which field in Roon itself can be populated with such a Tag without having to do this manually.
Is there one?
Where do I sign ?
That’s sad, isn’t it; but something we’re used to.
Your help, clarification and support very much appreciated, Tony!
Your composition is still not identified properly. Roon was not able to match your composition with the allmusic canonical form. You should be seeing +/- 70 performances from your streaming partner, not 4. If you click on the library or disk icons you should land on this page:
The quickest fix is just to use the allmusic canonical form for the WORK tag as I suggested. None of your 4 compositions are properly identified.
Your album is not identified either. Album identification will usually retrieve the album art and auto-populate the Classical genre tag. You seem to have figured out how to add a Classical genre tag manually. Otherwise I don’t really understand your question about adding a Classical genre tag. Its just the way you have done it. You can also add it in batch mode by highlighting your albums and going to the album editor. I only ever do that as a quick fix. It is better to get into the habit of tagging your source files with Yate instead.
I may have been labouring under a misapprehension.
Namely this: am I trying to get from AllMusic an ‘intersection’ which captures both work and performance - and Label number for good measure where there are more than one performances by the same performers?
Or am I only retrieving just the canonical/accepted record for a performance, a work’s name from AllMusic?
The idea being then that Roon will consequently know where to get its ‘many known performances’ (the disk icon with which we began this thread) and thus be able to populate that field in its own database?
And this in no way compromises the records of the downloaded Album - e.g. from Presto etc?
Is it akin to correcting a spelling?
And are we to conclude that it’s only because that one field (Schoenberg’s work’s title) failed to conform to AllMusic’s ‘model’ that everything else was ‘off’ in Roon?
Yes, you can look at it that way. It is correcting the spelling so that roon will match your performances with its internal representation of the composition. Roon can then count correctly the Qobuz and local versions of the composition that you have.
You seem to have higher expectations of roon or you are over thinking this. A roon failure to identify compositions is rather common, especially with unidentified albums with poor on-line metadata as in this case. So you need to manually intervene if you want a different result. That’s all. Your 4 “compositions” are only linked to each other. They are not linked to the universe of recorded performances. You need to manually fix that to get the real benefits of using roon.
Unfortunately that Gielen Vol. 8 box-set seems to have terrible metadata and will require a great deal of manual intervention depending how much time you are prepared to devote to it.
I understand and follow everything you say. Thanks.
GIGO: if the metadata is poor online, I can see how Roon cannot be expected to ‘improve’ on it .
Is there any way to know (for sure) when Roon has ‘failed’ to make an accurate match with AllMusic?
I’m already planning it. I shall split up the 200 Albums which I have stored in Roon, accumulated over three years or so, and edit them in batches. A long job. But my uses, as you say, will be the better for it.
What’s been misleading me, I think, is an assumption from posts like this one of mine three years ago (in which you were good enough to help as well, Tony, thanks!) that Roon could do more than I thought it could: “Just import a file and Roon will do the rest”.
I have a lot invested in Roon and want to make the most of it.
Your help, guidance and suggestions: INVALUABLE. Thanks so much again .
.
