Evaluation of a Roon launch with millions of music titles

@wizardofoz Yes, that’s how I see it too. Music has made us grow as a community and if it doesn’t play, we won’t be any happier. There are so many ways to find beautiful music. If you’re limited to the internet, you’re making it harder than it needs to be.

My local files are around 300K tracks…I’ll survive a few months without the internet :wink: add to that unripped CD’s SACD’s and a good stack of vinyl and I’ll be just fine too.

2 Likes

Great love can also withstand crises. Roon will solve it.

…and the music plays with Roon

This is why the rest of us have a slow Roon experience :wink:

3 Likes

I don’t think that our patience makes Roon slower. Only our impatience or hectic actions lead to errors. Well-considered we come to the good future and every advice should be checked by ourselves.

Use DNSv4 servers assigned by the ISP (recommended).

ISP German Telekom is the same

(changing to Google / Cloudflare has no effects)

Screenshot_20220606_073917

Temporary Lollypop without Wine and Roon with PlayOnLinux

Roon has also resumed work here after many technical difficulties. Therefore, I will not hide the daily balance. 1% of Tracks last 5 Hours

Screenshot_20220606_200143

Since there were also questions about alternatives here, I used the time to get an updated overview of what Lollypop can do in version 1.4.33-1.


Fancy media playback integrated into the system, good full screen.





Many thousands of texts and images are found at megabyte speed, more than Roon because there are more sources. Few misses and clever maneuvers when an artist’s image is hard to come by. With Roon rather few kilobytes come through the same network, because the other side has to think and share more, my system remains underloaded.

My conclusion, it can look a lot fancier than other programs, share a lot more images and text, but it will never work with my library. Just reading in 1000 or 10000 music tracks is a joy. It also comes with a lot more pictures, lyrics, Wiki and other information…Spotify is on board…so much further than many others and yet still a good distance away from what our beloved Roon accomplishes when we find connection.

The last two hours were really good again.

Screenshot_20220606_221248

Impressive collection @Uwe_Albrecht. I’m currently at ~170.000 tracks. Don’t think Roon is created for millions of tracks, but I do hope your work here will lead to a more optimized Roon for us all :slight_smile:

@Tor_Gunnar_Berland I am deeply convinced after my excursions in many music services that AI will help us to find the music we like most.

Man is so resourceful and no longer walks every travel distance. The traveler uses the horse, a bicycle, the horse-drawn carriage, the automobile, and then sits on a plane and travels the world in a few hours. What can these people tell? AI is the travel tool of the next century here, but Valence is still delicate and small. However, she shows me the musical world of the 1920s, which was played in Berlin as well as compositions that were created in this place only this year. Genres from swing, jazz, opera, old hits, large orchestra to new German wave, krautrock, new German hardness or lo-fi hip hop in Germany come to light, where old crackling and cracking, old styles and instruments are culturally repackaged for the modern age and then spread rapidly in the music scene all over the world. When Jazz Friend reads about a rapper, he keeps scrolling. Is that right? Here’s an example:

to get there you need this:TIDAL
and this TIDAL
MF Eistee
MF Eistee Picture from Discogs

A contribution of the Bayrischer Rundfunk about it:

or further technology there (selct what you want): 2021 Around the World

with my selection German / lo-fi jazzhop 2021 Around the World

Valence doesn’t want to be like the previous leading technologies (EchoNest, Niland…) and make the selection even more democratic, qualitative and just extensive enough to really discover the whole music world with pictures and lyrics.

If a genre, a music style or an artist can never be found or described, it’s not a good journey into the whole world of music. Every single data point should be a convenient and quick journey into the interesting world of music.

If only the Rolling Stones or the Beatles are found it is just as poor in popular music as knowing only Beethoven or Mozart in classical music. This does not cover the smallest linguistic and cultural area, even if you like to listen to it all very much. Roon is made for people who want more and are happy to pay for it.

We can only collect, enjoy and share our experiences of how far it has come.

Those who really want to help also share brittle statistics, technical alternatives, mistakes, problems and a lot of fun to keep the mood good.

06/07 06:14:23 Info: Starting RoonServer v1.8 (build 952) stable on linuxx64

Today in the last 14 hours 78715 pieces of music have come through the second step identification. Thus the still open part sinks under the 1.3 million mark and I can show that more than half is done.

Screenshot_20220607_201939

The second picture shows that it went forward to the previous day on both drives. A total of 80265 albums are now identified. I estimated that Roon will work on this task for more than two weeks and will be able to identify about 160000 of my 323549 albums in the end.

I am always up for positive surprises. The development team is continuously working on improvements and fixes and surely if everything works out they have built a unique, but not yet perfect solution for huge databases.

sooloosgang

With this photo, I want to bring back old memories of CES 2007.

Enno Vandermeer and his team already showed interesting products back then and explained further plans. At that time the company was still called Sooloos.

Too bad that despite my love for music, I did not discover and use these products before 2020.

I’m not convinced. As long as machine learning knows nothing about the underlying music itself all it is able to do is crowdsource the lowest common denominator. Through analysing the underlying audio MusicIP understands things about the music itself and to this day produces far better playlists than anything else I’ve tried. Nothing else comes remotely within firing distance of what MiP musters.

2 Likes

Articles like this interest me and it’s AI. Many reference books are written about what can make good AI a successful recommendation system. But there are now more research papers for commercial purposes that are not shared.

The best system certainly no longer uses just acoustic fingerprinting (there are several approaches to that as well), but much more secret knowledge. Even this blog is part of the AI. The many messages to LastFM…

Most data of 400000000 customers and extremely expensive acquisitions still ensure Spotify’s technological lead. They’re smirking about MusicIP.

But I find it interesting that my first paid music service MusicMatch is mentioned in the above article as a spiritual father. They were really far ahead at the turn of the millennium and were flattened by Steve Jobs.

What a precise music analysis can give today, you can test rudimentary e.g. on this page with your own Spotifyplaylists:

Of course, not the whole AI knowledge comes to light, but a small part. This used to be research work at EchoNest.

You can even choose how much you want to see.
Screenshot_20220608_082900
For me it would be a dream if Roon would approach this and expand it with own ideas. Each song should be linked with one click (link) to all legal play sources and always available, then the local collecting is pointless, but not over. It is then only a special rare way to show love and save.

Yip, used it for a while after MiP was killed off before the the Echonest API was shut down to drive Spotify. It was good, but it wasn’t MiP. The kind of exploration and discovery you write of is precisely what MiP was/is superb at.

1 Like

MusicIP had the starting advantage of structuring preferences limited to your library. Today’s AI has to cope with Soundcloud 300000000 tracks and pack them into suitable song packages. Again, you have to feed Spotify first so the AI can work out what suits you.

Add your favorite local artist tracks to Spotify in a 500 song pack using TuneMyMusic and have it create song radios or artist radios for you. For me it’s a dream and mix I can alone.

Is it just me that want to discover new music by myself, by interest and curiosity? Why would i want to be force fed with what “the commoners” like, or just being handed everything without personal effort? AI will make you an idiot! :wink:

2 Likes

I’m already an idiot for most platform operators because I collect so much and don’t just want to stream. :joy: :rofl:

I only use AI for my journey of discovery and it is very personal, obscure and better because I can couple my own thoughts and actions with AI. This works even better than handpicked from supposed or actual experts in the public domain. Every radio station plays the same stuff. That works differently for me with AI!

Here in the forum there are a few for jazz and classical that still pass on a personally better experience 1:1, but it must additionally meet my taste. I can then perhaps see how high quality (technically speaking) the recommendation is. I can shine with better knowledge, but can I enjoy the harmonies, the vocals, the instrumentation, the recording technique, the musical style, the technical sophistication and much more in a relaxed and lasting way? It’s another excursion into a different world of music that I like to do and then use AI to amplify what I like.

Hence the many starting points almost always off the beaten track, but your socialization and listening habit is the bedrock of any (AI) expansion.

At the moment it is mainstream that I am an idiot with this thinking, but I don’t give a damn about that.

More advantageously described:

Mikael and Uwe go their own exhausting ways and yet are connected via Roon.

1 Like

Spotify and their ilk drive music AI to favour to the lowest common denominator. It’s been reported that 24% Spotify users skip a new track within the first 25 seconds.

Songs are getting shorter and more similar in what has been dubbed Spotifycore.

The platforms are conditioning users to select what the AI serves them and the users bias the AI in a feedback loop. I’m not convinced that AI will select music that will challenge me or bring me joy as a listener.

1 Like

Paul Lamere has played a major role in EchoNest development and his blog is legendary. The knowledge and the article is from 2014 and the research work on this is even 10 years older. There are two streams of thought on this topic here in Germany.

The public broadcasters who are in competition naturally see only the dangers and wrong uses. Here is an example:

Netflix & Co. - Entmündigung im Streaming? | deutschlandfunk.de services%20are%20the%20crisis%20winners%20of,%E2%80%9C%2C%20said%20Kleiner%20in%20Dlf.

But in fact they already cooperate with the competition offer especially podcast, live events, cultural programs etc.

Those who want to look at it a little more neutrally will find this, among other things:

https://www.zeit.de/kultur/2021-05/streaming-dienste-auswahl-serien-filme-algorithmus-kritik-netflix-mubi

https://www.delamar.de/musikbusiness/spotify-kritik-60248/

In the past, public broadcasting was a lonely power in cultural events and for the formation of fan groups (creation of new idols) in a unique position. But this necessary cultural work must not stop the progress and democratization of culture! I am not one of the critics. Traditional radio plus streaming used with personal expertise is just great for customers and the financial future for artists and publishers. We can talk about revenue or more equitable distributions. It’s harder than the lobbyists tell us it is. Even concert scheduling and much more is now done by AI from Spotify.
Finally, not only mainstream is played when the intelligence is not only programmed smartly in the computer, but also sits as a smart head in front of the computer. Yes, it is said that drivers bluntly drive four times into the same road works because the navigation device dictates so. The feedback loop, the song length and other observations are rather a personal shortcoming in not sensible use.

Is the offer known? https://www.midiaresearch.com/

I still swear by my album cover selection methodology - if the album cover is interesting I’ll have a listen, if not I’ll pass. Have discovered many new artists and great albums this way. MiP’s magic was in weaving together songs from across your entire collection in a way that made for great listening experiences that could match mood or type of music you want to hear.

1 Like