You can’t be telling me that it is beyond the wit of man to push recordings with less than 4 minutes of music by Beethoven lower down the list of “Most Popular” than third place. It really isn’t difficult.
we all use (and love) Roon for your very same reason(s) so we are complaining about how those things changed from 1.7 and are now more clicks away, changed to something less effective and/or poorly performing because of some changes
then there’s the massive waste of UI space on anything other than a phone/tablet 
2 Likes
At cloud scale you have so much information you can do more than simply indexing.
My favorite memory: all my 45 years in the industry people have tried to do human language translation with dictionaries and grammatical rules and failed. But after the Haiti earthquake, Microsoft Research used an AI system, trained it on vast volumes of internet data, and created a bidirectional translator between English and Haitian Creole, for the benefit of the international aid staff, in 6 days and 7 hours. That is not possible from one hard drive.
Yes, Roon can index my local library, but it can be so much smarter with streaming services.
That’s why I think 1.8 represents an essential innovation. It provides the common paradigm for exploring and discovering music across both local files and streaming services.
This is essential because it is obviously wrong to suggest that we will move to a future where streaming will take over. That transition has already happened.
It is to Roon’s credit, and my intense gratification, that I can browse and play my eclectic local library and the streaming services without having to know the difference.
1 Like
„Pretty weak“ is putting it very mildly. If Roon tells you a Trompeter and a Mandolin player are among the 4 or 5 „Top Performers“ of Chopin‘s piano works, then you know something is seriously broken.
1 Like
I disagree. That’s a unique use case with classical composers. Machine learning will always have gaps in it but you’d need a lot more wrong cases across genres to say it’s seriously broken. Most of it seems fairly accurate even if it’s not the most deep, in the pocket pick.
Please remember how the Roon Management has been bragging about all the improvements made to the „unique use case“ - as you call it - „Classical Music“. So where are they? Is this all they „improved“? Trumpeters performing Chopin Preludes?
I agree entirely. Data at scale can solve problems we used to think were insurmountable. And I have no issue with integration with streaming services, as long as it doesn’t come at the cost of local playback.
As a side note, while online translation is great, I suspect If you actually knew both languages well you’d find many of the translations rather ridiculous. There’s a difference between asking if you’ve eaten and discussing the relative merits of different recording of Beethoven.
@Jez totally it’s doable, but people often forget that their simple fixable pet peeve lies amongst thousands of others and every decision is a compromise.
It’s not my pet peeve, and I do realise every decision is a compromise and all sorts of other platitudes. I was just suggesting changing the construct of “discovery”, for which something produced by the roll of a dice could fit the bill, to “curation”, which implies being looked after, cared for, or in the words of one online definition “ the selection, organization, and presentation of online content, merchandise, information, etc., typically using professional or expert knowledge.”.
1 Like
They bragged about all the valence functionality. Which isn’t even there if you aren’t a streamer. So yes, 1.8 underdelivers on the marketing in a major way. But I could say the same thing about filtering still not working with over 5,000 albums in my library too (2 years after breaking their filtering).
Search and discovery both need more work. I’m just saying it’s not as bad as your specific case which is the most complicated case with missing classical metadata in so many cases
1 Like
The bragged about classical just as much as bout Valence. I will look up that bragging and post in here a bit later.
Rotten tomatoes for music? Or some other meta critic.
My first thought is there just isn’t enough volume.
That’s the problem. They didn’t mention - even in the demo - that most of the improvements are coming from streaming data, even with classical.
The whole presentation of 1.8 felt very bait and switch when I jumped in. Like, “what did they actually give me that’s new besides the skin?”
1 Like
Et voila! We‘re getting on the same hymn sheet now. 
1 Like
I would imagine that Roon earns revenue from streaming sites when users sign up. If they don’t I’m sure they will. So that’s going to shape the development of the app.
What not spin it off? Roon Library vs Roon Streaming?
as promised, here the statement of @brian, CTO of Roon from 10 days ago.
3 paragraphs of „bragging“ about how they improved classical Music
Where is the positive impact of all this on my library? I fail to see it.
1 Like
I only have 195 classical albums in my library but it did increase that from 164 recognized under the genre before.
Signed into my null Qoboz account, I get a fairly good selection of recommended albums, new releases that seem to correspond to my listening (ie a Max Richter release who is one of my most listened to modern classical artists).
So on my end the basics look to have improved the genre’s visibility and connections for me.
My inclination is that if you’d done a lot of work with your classical content to make it more identifiable to yourself in previous releases, these aspects may not be as noticeable. The case you mentioned though was a specifically bad album/artist recommendation with Chris Thile who has a slight connection with some compositions he’s played on but is obviously not a seemingly relevant connection.
Is that Roon trying to direct you to an adjacent genre and composer/artist (a string master)? Or is that Roon shoving an oddball rec your way?
In my case above, Richter was surrounded by a Beethoven Sonata and some other more traditional pieces to check out. So it seemed to give some diversity between classical and modern classical. A 15% jump in identified classical music is an improvement for me, but that’s about .5% of the music I have in my library so it may be drawing more out due to this being a weak point in my library.
Anyhow, I just said I didn’t think the valence recommendations were terrible. If you’re seeing a bunch of issues specifically with your classical music I’m not as adept in, you should put them together and start a thread to give the classical use cases more visibility to the Roon team. I’m glad to look a little deeper into these features with you as “backup” because better connections and AI should help all of us, especially with unique libraries.
Total sidebar… as with many other things that have a binary “like” or “add” (such as the “add this streaming track to my library) I do wish there was a 2-step or even three step that allowed me to say “this is part of my desert island discs”. I do realize that’s what some people do with tags, I’m just not that organized. So I imagine a “like it even more” and “no really I mean I love it”. But I may be a segment of 1 here.
Another example of the benefits of this integration, I just posted: