Live data from Hacker News

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

maroofy.com

121–130 of 444 posts

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

#124
post #24

This is very interesting, but unfortunately I haven't had the greatest luck in finding new songs I would enjoy listening to. It absolutely finds similar sounding tracks, but it doesn't distinguish which part of the song made it enjoyable. There's no tempo consistency or genre consistency or even main instrument/vocal timbre consistency between recommendations. I think locking one or more of those dimensions would all…

I am enjoying Raga's Dance, which is nothing like what I was just listening to. Thank you for the recommendation ;)

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

#125
The search feature is great from what I can tell!

I tried an obscure(?) Japanese rock band "9 Ball" and they were extremely high on the search results. I can't even find the album by using Apple Music.

Searching for 9 Ball in Maroofy shows them in spots 2-6, easily: https://maroofy.com/songs/185487468

When I try to search for the same band in Apple Music, I get this jumbled mess of results: https://music.apple.com/us/search?term=9%20ball

Even searching for a specific song of the band, the correct artist is 5th on the list: https://music.apple.com/us/search?term=9%20ball%2024%20hours and doesn't even link to any of their songs: https://music.apple.com/us/artist/9-ball/28599994

I literally cannot use Apple Music search to get to the page to buy Sound Seeds by 9 Ball, without going through Maroofy first. I had thought Apple removed that album years ago due to some kind of licensing issue. Strangley enough, even though there is an option to "Buy for $7.99" clicking on the button in Apple Music doesn't seem to initate any kind of prompt. Maybe this is old data, but the previews are still available? Who knows. Either way, the search for that information was excellent.

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

#126
Searched for both "portishead" and "wandering star" and it just showed me songs or artists with those words in them, even if the music wasn't even remotely close ("lucero - wandering star" is very very different than "portishead - wandering star")

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

#127
I've found probably the most egregious example of what knaik94's comment is talking about: https://maroofy.com/songs/214977681 ("Being Alive" from Company)

All the AI model seems to understand is the opening 5 seconds piano. Listen to the actual song it just opens like that because it's from a play.

I think this will struggle with any song that has build up, it's very promising, though you need a sample of entire songs not the sample Itunes gives you.

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

#128
Seems like the recommendations are somewhat hit or miss so far. I would suggest finding a way to say why it recommended a particular song. I know that this is already common, but maybe you could come up with a new spin on it.

Also, other comments have suggested that this was trained on 120MM of the audio previews instead of the full songs. That might explain why the recommendations seem a little off for some people.

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

#129

Feedback: "sounds like" for music is more than just rhythm and tone. The search results are all useless for me because two songs with similar sheet music often have wildly different lyrical styles and subjects.

It’s interesting from a musicology POV.

I was looking at Islands in the Stream by Dolly Parton and Kenny Rogers and it had some Latin neighbors that make some sense to me from the synths in IitS.

Post reply on HN