I was asking for it in a far more technical form than I think the average user might use. 7/8 time, major key, female vocalist? No problem! Synth, brass, and prominent banjo? No problem! But finding those that exist without using the genes to seed a station would be pretty tough, and even seeding a station with such tracks is no guarantee that it's going to recommend what you're really asking for. I want this, as a musician, to see if any wacky idea I have has been done, done well, and what worked or didn't work. It would also let me really zero in on stations I simply listen to normally, where I see a strong pattern of genes but the algorithm doesn't use those as an absolute. But I've been asking for this for a decade. I once got a reply from a dev on a blog or something that they'd bring it up in a meeting. (2013 or so? I forget.) Then this suggestion would suddenly have no mention or response anywhere, as if it wasn't just shot down, but actively suppressed. It's hard to generate revenue outside of ads when the user can point the algorithm far away from popular music that a label might pay to promote. The fact that the back end for this basically exists already, and we'd only need a UI has me convinced that we've never received this feature because of money, and there's no chance they tell us whose money keeps it away.
... View more