Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: I trained an AI model on 120M+ songs from iTunes

maroofy.com

Post reply on HN
Print view View original

431–440 of 444 posts

Re: Show HN: I trained an AI model on 120M+ songs from iTunes

#431

Post by achairapart »

huhtenberg wrote: post #306

Search doesn't work in a recent Firefox on Windows - typing anything in the search box shows the "Loading..." dropout below it and then nothing happens. This generates no network activity and the console shows: Cross-Origin Request Blocked: The Same Origin Policy disallows reading the remote resource at https://cdn.segment.com/v1/projects/F4GFNelOpRsgUJc6iwTuiXr2t6AH5LCY/settings. (Reason: CORS request did not succee…

Same. Get stuck on an infinite loading. Even without uBlock and similar. Too bad sounded cool. :(

Re: Show HN: I trained an AI model on 120M+ songs from iTunes

#433

Post by joshu »

subtech wrote: post #229
joshu wrote: post #73

this is pretty awesome.

<3

Given song to song relationships, can you aggregate to artist to artist relationships? For all songs by a given artist, get all the relationships, and what artists are highly represented?

Re: Show HN: I trained an AI model on 120M+ songs from iTunes

#434

Post by raffraffraff »

class4behavior wrote: post #324
raffraffraff wrote: post #261

Earlier quoted context omitted.

Which is unfortunate because it has (on a tiny number of releases) instruments and vocal tags. It's just so unreliable. AllMusic is another decent source for tags, but not instruments. It's the age-old problem with ML/AI: data quality. Garbage in, garbage out. If only we could crowd-sourcev listeners and get them to tag music from a list of available moods, instruments etc. Oh wait .. that's exactly the feature that…

rateyourmusic.com is an alternative to allmusic.com while chosic.com focuses on finding music. The later has generated more similar music for me so far. But I welcome every additional project improving the search for music which has been so neglected by most services.

I've been a paying subscriber to rateyourmusic for years because one extra feature you get is per-track ratings. Extremely useful if you're in discovery mode.

Re: Show HN: I trained an AI model on 120M+ songs from iTunes

#435

Post by mr-pink »

ojo-rojo wrote: post #62

I've never commented on HN before, but I feel compelled. Congrats, you get my first. I've been using your site for 10 minutes and already added song after song to my library. I'm sure you'll improve the matching algorithm over time. This is a great first step; I'm being exposed to songs I've never come across before. Good job!

good job to you for commenting. i wouldn't have checked it out otherwise

Re: Show HN: I trained an AI model on 120M+ songs from iTunes

#436

Post by pluxx »

Fascinating stuff! DJing mainly electronic music, I tend to base a lot of my match-making on beat and rhythm. From playing with this AI model a little, I feel there is more priority placed on timbre than transient information? Really looking forward to seeing where this goes!

Re: Show HN: I trained an AI model on 120M+ songs from iTunes

#437

Post by efreak »

huhtenberg wrote: post #306

Search doesn't work in a recent Firefox on Windows - typing anything in the search box shows the "Loading..." dropout below it and then nothing happens. This generates no network activity and the console shows: Cross-Origin Request Blocked: The Same Origin Policy disallows reading the remote resource at https://cdn.segment.com/v1/projects/F4GFNelOpRsgUJc6iwTuiXr2t6AH5LCY/settings. (Reason: CORS request did not succee…

Stuck on my mobile device. Disabled both ublock and Firefox enhanced tracking restrictions. Tried brave, chrome, Samsung internet, and regular Firefox (usually use Mull). Nothing works.

Re: Show HN: I trained an AI model on 120M+ songs from iTunes

#438

Post by 1Y3 »

Super interesting to read the varying opinions on here. Seems it differs a lot for music styles and personal preferences. I tried a few songs from my favorites and while the similarities were cool to see, it didn't correlate to actually liking the song for me that well. Then again I couldn't say what makes me like a song. My best guess currently is a correlation the voice of the singer(s) for songs with lyrics but not the text in itself.

Still really excited to see updates on this, especially recommendations based on a playlist, since the big 2 (Apple Music, Spotify) haven't figured out recommendation for a all-over-the-place music taste yet.

Re: Show HN: I trained an AI model on 120M+ songs from iTunes

#440

Post by efreak »

popinman322 wrote: post #118
paxys wrote: post #108

Solid work, but this only reinforces my belief that music recommendations is an area that is not (yet) cracked by AI. You can find similar songs with great accuracy, sure, but even if two songs are close to each other by all calculable metrics they can still vary drastically in subjective quality. And a quick suggestion – all my top results when searching for popular songs were songs that were either different versio…

Have you tried Gnoosic[0] yet? It's a more standard recommender system, I think. Some of the recommendations are really on point. [0]: https://www.gnoosic.com/

Gnoosic and other existing tools are entirely useless for some things. Most music tools don't even allow entry of individual songs, which can really mess things up if your artist has changed sound styles over the years, or plays in multiple genres.

For example, any time I hear music I like in a Disney movie and want to hear similar music, all I can get recommendations for is movie soundtracks or whatever the guest artist is known for.

The example that covers up for me on occasion is the soundtrack to The Road To El Dorado, specifically the instrumental tracks (Cheldorado, The Brig, Wonders of the New World); Chosic.com gives me a ton of other soundtracks, mostly orchestral scores, for things like Kung Fu Panda, LOTR, Assassin's Creed, Angry Birds, POTC, etc. While it's nice to be able to tweak the algorithm on chosic to pick quiet movie scores for background music, that's not what I'm looking for. And gnoosic is entirely useless for that.

Post reply on HN