I think that only works if the song is in the queue with other tracks after it in the queue. Plus it’s many clicks. There is no reason why starting radio from the currently playing track shouldn’t be a single click, yes?
ahh, I just saw the reference to the post about hitting the blue “end” button. I will try that. But not really intuitive? End?
Great when it works but pretty frustrating experience so far. Some talk upthread about the difficulty with edge cases, in this case hip-hop. Seems to me anything but well known classics will quickly grey out and stop recommending. First, it’s not clear why radio is stopping. If it’s a known thing that obscure seeds will stall radio, maybe there should be an error message along those lines.
Second, it’s not even an obscure track, it’s DNA by Kendrick Lamar, a well known track from the latest album by arguably the greatest rapper of our generation. So…musical taste aside, how would I go about finding related material? I’d start with the label. If the algorithm were aware of labels it would know TDE labelmates like Schoolboy Q, ab-Soul, SZA, Jay Rock, it would know producers like Terrace Martin, Tyler, the Creator and on and on. For example, Tidal has 100+ guest spots from Kendrick Lamar and not one of those has come up. Just my thoughts.
This, I have no way of knowing why radio fails to recommend anything, so this could all be unrelated. It’s still true though. Cheers
Ha, no surprise we share the same frustration here @lorin.
I’ve had so many of these dead ends with HipHop, even with the popular stuff like you mention (popular outside of the Roon world, out in the wider world).
Roon Radio is performing much better with Jazz, older Rock and Classical (I sometimes listen to these too but less).
A few of my friends have been waiting to see/hear how this new Roon Radio thing performs before committing to a Roon membership but after using it at my place and seeing how badly it performs with newer and more popular music, it’s a deal breaker (for them - I’m a Roon lifetimer), just from the Discovery aspect.
They’re not fond of needing to have three subscriptions, Roon+Tidal alongside Spotify/Apple Music.
We tested Roon Radio this past week and I tried to parrot how advanced this algorithm is but when we got dead ends with Radio, my friends said you can talk all you like about how advanced you think the algorithm is but the music has stopped, so it’s sh!t… I didn’t have a good response apart from, let’s give it more time. The same with artist’s heads cut off on Chromecast display. I tried to parrot that it’s using some advanced algorithm but again the feedback was, “great, but again it looks sh!t”. Again, I said let’s give it more time (will revisit in a few weeks - it’s been said something is happening in the background to improve this).
It’s still very early days of course (only 1 week since v1.6) and they’re quite open to re-trying later on but if the Roon user data for this kind of music is low, the experience may not improve as quickly as I’d/they’d like. Time will tell of course.
I’ve recently re-started a Spotify and Apple Music membership just to see how far things have come from a discovery aspect (it’s been 2+ years since I used them) and I’m really amazed at how much better and personalised the experience is. Even the Tidal app. Of course their subscriber numbers are at 100+ million who are probably closer to my age than the bulk of Roon users (just guessing).
That’s my experience, just 1 week into v1.6. It’s very clearly early days of course.
I’d just like to add that some kind of label awareness would naturally apply to many genres! Pavement is 90s indie rock but selections somewhat favouring Matador labelmates would make for, IMO, a much more curated sounding playlist. One example of many.
Thanks @andybob. Do you think it will be possible for the user in a near future to specify not to listen to pieces older than two years or to specify elements like this? I would be surprised that the algorithms could not do it. It is because I would like sometimes not to hear songs older than two or three years or to be able to specify this kind of precisions (it could be the opposite : songs just of 2002 or before 1992). For the pleasure but it could be also useful for kind of research or study.
I suspect that kind of user controlled listening will remain the domain of the Focus tools and in Library Shuffle play. Roon Radio is designed to be automatic. But if it is something that you have a strong opinion about, open a Feature Request thread and explain why you think that kind of capability would be cool.
@brian, any chance you’d consider implementing this in addition to / alongside the machine learning Roon Radio? That’d improve Radio for those not using streaming services and if it’s anywhere near as good as MusicIP’s engine was before Gracenote consumed it I expect it’d be a great addition to Roon’s Radio capability. It should also go a long way to addressing music styles/genres where Roon’s customer base yields a shallow pool from which to pick tracks e.g hip-hop.
We are planning to do something with audio feature extraction in the future.
To be clear, low data areas will be much smaller as time goes on. The data will show up + improve on its own faster than we can try to desperately squeeze more out of what we have. We are slated to triple the data set size by year end.
The main thing that audio feature extraction helps with is cold start (figuring out quickly where new content belongs). The other thing it might help with is in creating models for things like mood, energy level, vocalness, instrumentalness, valence, … which could be turned into yet another vector space in which to consider similarity.
No concrete plans to apply this to library radio unless something radical changes about how people use Roon. Building and testing a new radio algorithm (library or not) is a huge undertaking. Putting that kind of effort into maintaining two of them makes little sense when the global algorithm is the one seeing most of the usage volume.
fwiw I have deleted and re-built my library several times (I only use TIDAL, no files).
But I suppose you could say I have a “mental library” – certain artists / albums / tracks that I know and love, and can use as a starting point for searching from TIDAL.
My biggest challenge right now is dealing with “candidates” for inclusion to my library. TIDAL has a bunch of new releases and recommendations that I’m sort of tangentially interested in, without knowing much (or sometimes anything) about them.
How do I handle those? Add them to my library and tag them? Add them to my library and hide them? Add candidates to library, favorite the albums I know & like, and tag my “super favorites” ? I’m continuing to work this out (and maybe it’s a question for How do you Roon?)
Sometimes I see stuff in there that I want to delete, and I do.
But–I don’t mind having excess stuff in my library that I don’t care about, because for me library inclusion is much more about accessibility than it is about collecting. I don’t feel total ownership over every piece of content in here. It is kind of like a personal streaming service–lots of content that I can access, but not all of which I feel ownership over.
For the stuff that I really own in an emotional sense, I try to favorite it. Our product doesn’t do a perfect job at encouraging favoriting, so I feel like I am swimming upstream when I do it, but I think it’s the right place, long term, for me to keep track of that stuff.
One of the annoying gaps with favorites (and tags, too) is that we don’t support favoriting or tagging out-of-library content. This is a historical detail…I wouldn’t design the product that way today. We’ll probably lift those restrictions at some point. If we supported tags out of library, you’d have a really easy todo list solution.
I can’t even begin to imagine how this would work. The interface between a radio algorithm and the rest of the system is absolutely massive–way too much surface area for an API, and with an obligation to correctly implement business logic (or risk bugs) that we could never outsource like that.
Sure. No, I meant an outboard volunteer radio extension – Roon API provides music analysis details (tags, beats-per-minute, chord changes, whatever features you calculate on a per-track basis) on tracks, and the extension would just tell Roon to play tracks. But the data exchange over the API would still be massive, so probably not.
This is how I do it too. Usually, if I’m vaguely interested in something, I add it to my library. Sometimes, I’ll return to it, and listen to the entire album. Sometimes, it’ll go untouched.
It’s interesting. I got in the habit of doing things this way because, if there was something I was vaguely interested in, I wanted to add it to my library, so that it would be available to Roon Radio. But, now that radio reaches into Tidal/Quobuz, this isn’t as much of a concern. So, the new radio functionality is changing how I think about whether or not to add things to my Library.
I have a relatively weak server PC, but it works well for now. One thing that keeps me from upgrading this piece of hardware is knowing that, if it starts to slow down, there are a lot of Tidal albums I can delete from my Library to reduce demand on the PC, because there is a lot in there that I don’t feel the need to “own”.
Yep agreed, and maybe this is just down to the personal nature of listening to music in general, and using Roon specifically.
Here are some examples from my recent Roon experience:
2001 by Dr. Dre – I’ve been listening to this album for 20 years. It doesn’t really matter if it’s in my library or not, because the mechanism I use to find it (search) is essentially the same whether it’s in my library or from TIDAL. Adding it to my library is more a matter of convenience than anything – to make sure I have easy access to the version I want, that it appears in shuffle / random, etc
Inner Monologue Part 1 by Julia Michaels – This showed up as a new release. I love listening to Julia Michaels, so I absolutely want to listen to this album soon (I already have, but that was my thought upon first seeing it). If it weren’t in my library, I would remember it as “that new album from Julia Michaels” and I’d be able to find it on TIDAL easily enough.
Heard It In a Past Life by Maggie Rogers – Another new release. I don’t know the artist, but the name is familiar, so I want to check it out. I did, I liked it, I want to listen to it again. If it weren’t in my library, I would remember it as “that new album I liked by that girl.” My best chance at finding it again would be to scroll through new releases and identify it by name / album cover, or my play history. If I couldn’t find it in there, it would be lost to me until the next time it got suggested to me (if ever).
Sam Cooke – A friend suggested I check him out recently. He’s got 22 albums and a bunch of appearances on TIDAL. Something I like about Roon is I can add all / a bunch of albums from him, focus on the pick tracks, and play them in chronological order. But, I don’t necessarily want to commit to having those in my library.
Yeah. I think I’m feeling some anxiety over how lots of unknown stuff in my library will affect the library radio experience. I suppose the best way to alleviate that is to use library radio, and if there’s stuff I don’t like I just remove it from the library
Yep… I often want to tag artists / albums without adding them to my library, so I can find them to check out later.
The trick would be, how well would it integrate with Focus? For stuff in my library, I can say “show me blues albums recommended by friends that I’ve never listened to” (genre: blues, tag: recommended by friend, played: never). I think it would be really cool to use that to find stuff from outside of my library, but I don’t know if that’s realistic.
(the most important part being that it’s not possible today, so my main focus is on finding things that I can actually do with Roon today)