Show HN: I trained an AI model on 120M+ songs from iTunes • Hacker News Board

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

#61

Post by brentis » Fri, Feb 03, 2023, 2:22 AM UTC

knaik94 wrote: 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 had the same experience. Could see the element which it matched with, beat, pitch, etc.but missed the riff or nuance that made the source song special to me

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

#62

Post by ojo-rojo » Fri, Feb 03, 2023, 2:22 AM UTC

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!

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

#63

Post by foysauce » Fri, Feb 03, 2023, 2:24 AM UTC

tastysandwich wrote: post #60

Personally for me the best way of finding new music is friends whose opinions I trust. I find going by similarity will pull up a lot of derivative artists that suck (eg Last.Fm/Pandora). I want something good, and if it's radically different, even better! One awesome use-case I can see for this though is finding alternatives to copyrighted songs. Let's say you make a sports video, and you have this fantastic song in…

I believe a very similar use case was the original primary function of Richard's compression algorithm in the show Silicon Valley. Which of course was entirely underutilizing the technology but that was the joke.

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

#64

Post by vivegi » Fri, Feb 03, 2023, 2:25 AM UTC

This is a very music recommendation engine. The best feature I guess would be the serendipitous finding of unknown artists we might not hear otherwise. Great job!

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

#65

Post by bfung » Fri, Feb 03, 2023, 2:26 AM UTC

Good idea - can use work on data cleaning though; got a lot of duplicate listings for some songs searched.

It’s like feeding Apple Music through Shazam - wonder why Shazam doesn’t do what you’ve done here

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

#66

Post by Izkata » Fri, Feb 03, 2023, 2:27 AM UTC

charcircuit wrote: post #52

snow_mac wrote: post #49

How did you get access to 120,000,000 songs from iTunes? Not just the listing, but the actual audio.

The preview audio is free. It may just be 120,000,000 previews.

Totally explains this part of the current top comment here:

> but much more often it feels like a 20 second section was used to define the original song and it misses the underlying concept

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

#67

Post by fallingmeat » Fri, Feb 03, 2023, 2:27 AM UTC

snow_mac wrote: post #49

How did you get access to 120,000,000 songs from iTunes? Not just the listing, but the actual audio.

im also curious

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

#68

Post by hasbot » Fri, Feb 03, 2023, 2:27 AM UTC

If I'm listening to a song, I don't want the next song to sound similar, and the next, and the next. For example, my playlist just played "I touch myself", "Whiskey in the Jar," and now "Somebody that I used to know." All of which sound entirely different. But, maybe I'm an outlier. Have fun!

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

#69

Post by fallingmeat » Fri, Feb 03, 2023, 2:28 AM UTC

paxys wrote: post #41

How the hell do you even get access to the entire iTunes catalog?

i think that would be an even more interest post.

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

#70

Post by Izkata » Fri, Feb 03, 2023, 2:31 AM UTC

jaimex2 wrote: post #48

Pandora?

Yeah, I don't know if they changed it but this sounds similar to how Pandora used to do music recommendations: Models based on various aspects of the song, then find best fits to those aspects. It would even tell you why a given song was suggested.