There is an opportunity for Octave to surpass Roon in user-friendliness, besides SQ. I’ve spent some of my time off this past week on the tedious task of trying to fix the many metadata issues in Roon. World-class software should be able to use all of the data at its disposal to figure things out. Even demanding the standard artist/album file structure shouldn’t be mandatory if there’s sufficient file name/meta data info available to identity the track. Asking too much? I don’t think so given the huge collections many people have. Even the most basic tweaks file structure can take forever when multiplied by the album count. The system should be able to figure things out as much as possible based on all of the data it has available to it.
It boggles my mind that some of this isn’t working better by now in Roon, esp considering the price. It amazes me how Roon seems to be ignoring information right in front of it. Many of my main artists have their proper icon with an image, then a second generic one next to it with their name. The answer shouldn’t be that “I did it wrong.” If Roon knows the artist name, it should be in the right place. Also, I often needed to map songs where I was just moving a song title to its matching slot on the album. I wasn’t applying any thought in doing this to understand why it can’t be handled automatically. Also, something as simple as files named “artist-album title” breaks it, even if they’re in the file structure Roon wants. Clearly there will always be limits to figuring out this kind of stuff. But, I know that these things are realistic points to address. One nice thing is that programming this kind of stuff is infinitely simple and straightforward compared to anything related to SQ. I hope this kind of intelligence is what Octave is aiming for.
Actually, the most difficult part of Octave is the metadata and getting it right. Roon suffers from the same problem. It’s really hard to make Octave simple and easy to use. In fact, the simpler and friendlier we make it the harder it is to achieve.
Consider that the millions of tracks of music we have to catalog and supply metadata for are not organized with perfection. The sources of metadata are all over the map. Part of the challenge is to figure out what people actually want: do you type “The” to access The Beatles? Or Mozart to access Wolfgang? And If I type Wolfgang what do I mean when it’s not as easy as Mozart?
Or, cover art and artist art? There are hundreds of thousands of photos associated with artists - some work for the graphic layout of Octave, others don’t. You can easily wind up with a picture of someone’s pants instead of their face, depending on the photo. We’ve actually had to write complex AI heuristics to suss out the face and position the photo right. Sure, facial recognition seems easy but it’s not and that’s such a tiny part of the challenge.
It’s maddening when you look at the results of a music search for something that seems obvious to you and not apparent to the machine. Works of composers, movements of a symphony, search by orchestra, conductor, composer. Want to listen to music composed by Beethoven but played by orchestras Barenboim conducted and finding them on a greatest hits album?
It is anything but easy and continues to be a challenge. The truth is the software bit that plays the music is trivial to getting the UI right.
Paul, thanks for the details on what you are up against. Seems kind of obvious it isn’t as easy as we think, otherwise someone would have gotten it right already! By the tone of your post I’m thinking (uh-oh) that maybe this is a little harder than you thought? Might be a good idea to occasionally give us these “this is how it really is” talks so that the fidgety ones will back down for a while.
My comment was about data available on the users hard drive. I was mainly referring to the metadata associated with a music file when you look at a windows directory with the view style set to “Details.” Then there’s info based on the directory structure people set up. I think part of the challenge is intelligently applying both. Evidently some stuff in my “Clash” folder somehow identifies the band as “The Clash”, because this band is an example of Roon showing me these names next to each other in the Artist view. Regardless of the cause, these challenges result in the need for much easier tweaking of the outcome to correct such issues. An example of an outlier I would understand needing to correct is the bonus disk in my version of London Calling.
I can think of all kinds of other stuff I’d do if I were designing Octave.
No, it’s not harder than we thought though what can be said is that it is more complex. When we start a new project we never envision the degree of complexity - if we did we’d probably never start. We just roll our sleeves up and execute according to the roadmap. We’ve made huge progress. There’s an active GUI up and functioning. But, it still needs a lot of work. I will post its URL sometime early next year so you guys can play with it to see what we’re up to. Hang in there.
shaft234 said
Regardless of the cause, these challenges result in the need for much easier tweaking of the outcome to correct such issues.
Agreed. I also suggest flexibility and options the user can set for importing user data into the system. For example, do you want your folder structure imported, or do you want Octave to ignore it?
I can think of all kinds of other stuff I'd do if I were designing Octave.
Keep sharing, although I suggest starting a dedicated Octave thread if you truly want to put energy onto this. Or post in one of the existing Music Server/Octave threads, such as this one: click.
I honestly don’t know if this sort of thing can be done effectively for classical music, given that there is much more metadata involved than for popular music and that it is stored in a bewildering variety of ways.
Like many serious classical listeners, I spend a lot of time whenever I purchase a download or rip a CD to get the metadata into the format that I use. Even if the label supplies it accurately (something you can’t take for granted) there is still a lot of work to be done. I also use some custom FLAC tags that Octave probably won’t know what to do with but that I find very helpful in selecting music.
If your programmers are able to pull this off, it will be an amazing feat.
This is exactly why I believe Octave needs to have flexible importing. It should allow you to identify and import your custom tags, as well as import your other labeling exactly to your preference.
I also suggest that it allow the user to indicate what additional tags he would like Octave to add. For example, if you have not tagged conductor or have left many blank it would be great if the user could direct Octave to fill in the blank tags while leaving alone all tags with values assigned.
Another option is to import all user tags into a user database, and for Octave to create its own second database, with the user being able to select between the two.
employ the Agile development methodology, especially the 2 week sprint and user involvement part. In my experience, it is the best way to develop apps that are especially UI intensive. Enlist a small group of volunteers who will review the features of Octave as they are coming out of development and give you immediate feedback. You don’t need a full working product for this, it can be done via WebEx or any type of web meeting tool. Not only will you get valuable feedback on what you’ve done in the last two weeks but you may even get some useful ideas (crowdsourcing). You don’t need hundreds for this, a dozen or so users who are willing to spend the time, represent different use cases, and have different home setups will likely give you enough diversity in perspective.
make MusicBrainz at least one source for metadata. What I like about MusicBrainz is that if you don’t like it or see something wrong, you can submit edits. Also, you don’t have to solve every metadata problem. Things like where how to use the “The” in an artist name can be made user configurable (e.g., Minimserver) with a default that is acceptable to the majority of users.
I suggested Agile back in the days when PS Audio was working with eLyric. I don’t think this is common practice for developers of HiFi segment and needs some skills in this direction.
Agile isn’t something you can just switch to quickly. Any IT shop still using a waterfall approach needs time to change. The worst thing would be trying to change horses midstream. Since I’m still posting this late on Christmas Eve, here’s a record I discovered this morning:
I honestly don’t know if this sort of thing can be done effectively for classical music, given that there is much more metadata involved than for popular music and that it is stored in a bewildering variety of ways.
Like many serious classical listeners, I spend a lot of time whenever I purchase a download or rip a CD to get the metadata into the format that I use. Even if the label supplies it accurately (something you can’t take for granted) there is still a lot of work to be done. I also use some custom FLAC tags that Octave probably won’t know what to do with but that I find very helpful in selecting music.
If your programmers are able to pull this off, it will be an amazing feat.
To extend magister‘s comment:
I’d also really admire if PSA manages to improve this tagging topic, but to be honest, so far my opinion is, that there will always be manual work for those who have a special demand and for ripping CD‘s. When buying music online, the tags are already existing and quite good imo, by special demand, manual work is accepted in my case.
So for me these difficult tagging improvements are more a nice to have, than very important. There are many other features and GUI topics that would be much more relevant in my case. All these various demands on a library software and the impact in case of SW bugs make this project quite challenging imo. Also to fit all administrative functionality into a pure web interface.
Even if I think that streaming of local music (not streaming services) will never be for complete non IT affine folks, I think making it independent of a PC is the main great idea.
What I also noticed and why I think manual work will always be necessary is, that no tagging library like music brainz etc. I came across manages to recognize most special masterings (2xHD, Analogue Productions etc.), but mostly just the standard issues.
For an audiophile who takes care of best versions, this usually means retagging or switching cover art. So for me, what Paul already achieved (to preserve own tags) would be enough and it’s nice and welcome if even more would be possible, I’d just put most energy elsewhere.
I use Pure Vinyl for ripping LPs. One feature they’ve introduced in their latest beta is the ability to use Discogs for metadata lookup. The nice thing about Discogs is that they usually have just about every release for a title, including the latest remaster/reissues. They also indicate which country it is from (import for vinyl). Pure Vinyl starts by having you manual enter the LP catalog number and then takes you right to a page in Discogs for that catalog number (with CDs this can possibly be automated). Since there are sometimes multiple entries with the same catalog number, you can then scroll through and select the correct one, including cover art. Also, if you happen to get a new release that isn’t in Discogs, you can enter it, sort of like a wiki. That way, everyone after you can benefit from your edits.
I’d also really admire if PSA manages to improve this tagging topic, but to be honest, so far my opinion is, that there will always be manual work.
I agree; my point was not that manual editing can ever be avoided but that it will be incredibly difficult if Octave tries to make sense of classical metadata from an online source (beyond the most basic info; sometimes even that is difficult).
When buying music online, the tags are already existing and quite good imo
For popular music yes, but this has not been my experience with classical -- even from labels that specialize in that genre. It's getting better but there are still many gaps and of course different labels do things differently. One label I buy from gives only the last name of a composer. Another appends the composer's dates after the name, which I have to delete; otherwise I end up with two different composers, "Bach, J. S." and "Bach, J. S. (1685-1750)". If Octave looks up a CD from this label, will it understand that "Bach, J. S." is the same person as "Bach, J. S. (1685-1750" (who is the same as J. S. Bach and Johann Sebastian Bach and Bach, Johann Sebastian)? That's a real challenge.
There are many other features and GUI topics that would be much more relevant in my case.
I understand your point, but be careful what you wish for. Roon has a beautiful interface and many other good features but is useless because it ignores my tags.
Paul has said that Octave will honor the user’s existing tags. So Octave should be able to handle my music since my tagging is complete and consistent. Hopefully there will be a way to help Octave work with the user’s setup. One example: relatively recently an ID3 tag has been introduced for the name of a classical composition (“work” I think, but that’s off the top of my head). When I started with digital audio there was no such tag available so I set up one called “Composition.” If I can tell Octave that the FLAC “Composition” tag is equivalent to the ID3 “work” then things should go well.