Of course Valence still has a lot to learn too, but it’s gradually getting better for me. Is this true for all genres you love? I heard from our expert Andreas that even in CLASSIC, Roon is a real help to really discover what you like. But I also hear complaints that it’s not good enough yet. Maybe it’s also a question of whether we ourselves are doing anything to find the connecting lines that artificial intelligence needs to work reliably. In the past, I would have signed off on larger criticisms without hesitation, but these automated games speak a different language. It works here in Pop and is not at all boring for me.
I think those who aren’t looking for new artists every day and already know almost all of them will still discover something here.
When it comes to recommending the same genre I played, it’s very good for me and I have discovered so much new music. However it does start to repeat the same tracks after a rather short while and I have to skip a few until it seems to get back on track again and play new ones.
I’m going to let Roon-Valence run all day today and see when I get the first replay. So far 22 songs no artist and no album twice and yet everything fits.
It’s quite amazing, the database has 2.6 million tracks, surely half duplicates that are on different albums or samplers. I have edited about 100,000 music tracks in the last weeks with Foobar2000 (improved metadata, new folder structure) to challenge the database I have read so far and at the same time I am doing my Valence test today.
Roon is quiet in RAM/CPU load and makes absolutely no difficulties, just great joy that I can share here.
It’s been over 7 hours and Roon delivers the final songs for the Playlist-Valence100-Pop. It did not quite work out without repeats. The algorithm seems to have developed a preference for Howie and he wasn’t the only repeat in the range of artists and albums played, but it’s quite remarkable how artificial intelligence can support.