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

411–420 of 444 posts

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

#411

Post by yunwal »

rohith2506 wrote: post #409
m3kw9 wrote: post #406

Earlier quoted context omitted.

He’s not obligated to tell us if he knows it’s an edge he wants to keep

I don’t think I agree with that. If he is scraping, all he can say is yes. The details of the scraper is proprietary and that’s his edge for sure. If not and he found an unauthorized source of retrieving information, this reveals a serious security breach in iTunes API and it’s my valid concern as a paid customer. 120M is a huge number and it’s not even text. It’s media

I’m pretty sure the model is trained on the 30 second previews. If you type in a song with a lot of different sections, the suggestions match the preview and not the rest of the song. Bohemian rhapsody, for example

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

#413

Post by s3p »

piyush_soni wrote: post #187
nanidin wrote: post #180

Earlier quoted context omitted.

Apple Music subscription is $12/month and it allows full downloads of songs.

But aren't they DRMed? Or does it let you download in mp3 or related formats directly (I'm not on Apple music)?

You're correct, it's DRM only

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

#414

Post by mustacheemperor »

This is incredible! I've gotten really into some niche genres this year and in trying to discover more of it I have been really frustrated by Spotify's limited radio mixes - it seems like I've been "black holed" into the same 100 or so songs with rare exceptions, for any of the albums I autoplay from in a given genre. This really might be the best music discovery tool I've ever used, it reminds me of how magic Pandora felt the first time I used it.

My only request is to please let me multiselect some or all of the songs the algorithm finds and automatically create a playlist from them on Apple Music. The very first search I tried, for a song that Spotify always creates the same limited mix from, brought up a ton of music I want to check out.

Edit: on that note, it would also be neat to request a combined playlist that mixes together multiple searches. This might help provide feedback for the AI as well about what artists/songs people consider 'similar' to each other.

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

#416

Post by toomanyrichies »

I'm teaching myself deep learning at the moment, and learning about embedding vectors was the first "holy shit!" moment I had.

To me, it's fascinating that not only can you:

-represent things like words as vectors,

-map them in a multi-dimensional space, and

-use that space to find the "closest" neighbors (i.e. the most similar words)…

…but you can actually perform "mathematical" operations on them.

The canonical example is that, if you represent "king", "queen", "man", and "woman" as vectors in your embedding space, then you can ask your model "What is king - man + woman?" and (provided it's trained appropriately) it will return "queen".

I look forward to the day when we can ask something like "What is 'Bohemian Rhapsody' - 'Queen' + 'Velvet Underground'?". Which, if OP's model were to be trained on whole songs instead of previews, would probably be a reality!

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

#417

Post by mustacheemperor »

subtech wrote: post #33
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…

Hey thanks for the feedback! I definitely have a lot of improvement to do on the model, it currently performs better for some styles/genres of music than others. But the model architecture I'm using is kinda outdated as well, gotta iterate on it more to improve it further! I'm also thinking of letting users upvote/downvote results, which can also help improve quality on the ranking side.

I'll note my own experience, that Spotify and Apple Music both struggle to find me latin reggaeton outside a small subset of popular artists, and my first couple searches with this tool have found me so much music I've never heard before that matches exactly the 'vibe' I want to hear, and is introducing me to different-but-related sounds and artists I couldn't have found on my own.

I agree with the other commenter - this is huge for me. Please, do whatever you need to do to monetize this so it never goes away. I would love to pay you for this.

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

#418

Post by orange8 »

subtech wrote: post #281
orange8 wrote: post #266

The homepage is an empty page with just one search box in the middle. It does not work in firefox.

Hmm I actually use Firefox myself, but tbh, it can be a bit flaky atm due to the sudden surge in traffic. Should become more reliable soon!

It may be a firefox specific bug, as when I switched to chrome it worked well consistently. Maybe a network error due to origin policies?

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

#420

Post by subtech »

llagerlof wrote: post #419

Beautiful. Definetely could replace spotify "recommended" for me. Please, add some filters: - By language - By year (between, above, below) - By fame (so the user can filter out famous musics or unknown ones) Thank you!

Yup, working on it!
Post reply on HN