Hacker News new | past | comments | ask | show | jobs | submit login
Show HN: I built a site that lets users find playlists by songs they contain (dags.dev)
295 points by NomadicDaggy on Oct 10, 2022 | hide | past | favorite | 128 comments



I once built a site that used Spotify playlist data directly. The intention was that you could take one of your own playlist and find playlists that have the greatest overlaps. The intention was to find people with a similiar musical taste to yourself.

Interestingly I found a lot of new music that way and also found other music directions that suddenly interested me.

Thinking this a little further, it provides the perfekt basis for a dating application. Music is a very personally representation of oneself and if you find someone who has the same muscial taste, you already have a common basis for communication.

Obviously this never made it past developer status since Spotify does not like third party apps storing their data (i.e. playlists of other people). You can use the data but not store it or analyse it.

It made me think that Spotify should build a dating app since they would have no interest in keeping you in their dating app as thats not their main business, unlike other dating platforms that have little interest in losing their customers.


> Thinking this a little further, it provides the perfekt basis for a dating application. Music is a very personally representation of oneself and if you find someone who has the same muscial taste, you already have a common basis for communication.

I don't think a shared taste in music is that important, or at all, in a relationship but this is the premise of aptly named Tastebuds[1], a dating app based around favorite music. You can import your listening data from Last.fm and Spotify.

[1] https://tastebuds.fm/


In isolation, perhaps not, but musical taste plus another characteristic can say a lot about someone. (I might be biased because my future wife caught my attention when I overheard her say she liked a certain band in a group setting where most people hadn't heard of it.)

Spotify doesn't have its claws in real-life gatherings yet, but they might have something with their music (personality) and podcasts (interests, affiliations) data.


What does someone's music say about their personality?


If someone likes a lot of niche stuff and more complex, less accessible music, they are likely to take passion for music more seriously as a hobby than somoene who just listens to chart stuff (not that there's anything wrong with that per-say, but I feel being that level of passive would lead to incompatibility).

Certain genres and scenes are also linked to demographics and lifestyles (e.g. hyperpop being very queer leaning, chiptunes being nerdy, punk being political, bassline and techno being linked to the underground rave scene, etc...)

Also if one of your favourite things is going to gigs and festivals, producing music, or following it, having crossover in taste means being able to share a passion.

A large part of my relationship with my partner is listening to music, either at home on the hi-fi, or at events. If we didn't have a decent amount of crossover, this wouldn't really work.

Having exactly the same taste is probably not so great as I think the differences and gaps are a great space to grow and discover.


Openness to new things? Variety is the spice of life, how spicy are they?


True, there are definitely people who don't place any importance on music but for those that do, musical taste can be very important in a relationship.

So, no music taste isn't ultimate be-all-and-end-all factor but for some it can be very important!

Tastebuds seems only to have a facebook login - hm, not everyone uses FB! Strange that they don't have a Spotify login, probably the same issues that other apps face when they start to store Spotify data.


You can register by email and then import your info from Spotify.


You can base a relationship on plenty of other things, but shared interest in something is pretty important. Music can be a big one for a lot of people. My wife and I met at a KMFDM show, and traveling to music festivals is still the biggest social thing we have going on at all. Doing that without her would be a lot less gratifying than doing it with her.


I feel like there is a difference between sharing a love of music and sharing a love of the same music. I believe I would struggle with someone who loved no music and/or with whom there was zero overlap. But I'd not expect to love the same things


Somewhat agreed. Basically I found most people learn to like the music of their latest romantic partner.

I think partly it's just familiarity. There are plenty of "classics" I didn't really like when they came out but now they bring a feeling of nostalgia so even though they weren't "me" back then, for some reason they now feel like part of me, my history, my experience.


Back in the day we used last.fm (nee audioscrobbler.com) for this. :)

I have a largish group of friends that I met through similar tastes in music and being "neighbours" on last.fm.


It's nowhere near as good as it was in its hayday in 2006, but I wonder if, instead of specifically finding playlists with the song, you could instead grab 5 songs either side of it in scrobble histories. Sure, you'd have to filter out album tracklists, artist shuffles and then dedupe. But because of what last.fm is (99% a tool for listing everything you've ever listened to) it should have far more data to pick from, including all of your inspired ad-hoc mixes that would otherwise be lost.


We had the same idea about reddit years ago. Letting people opt into "dating" and then sending them matches of people who upvote similar articles. We were even building the recommender on the same technology. Not sure why we didn't do it, I guess we just didn't have the time.


Remember last.fm? If I remember correctly, it was able to show other profiles with similar music taste.


It still does - in the Neighbours view.


I still use last.fm, connected to my Spotify account. Before it was connected to the iTunes with their desktop app. I always loved their listening analysis and charts, matching with other people's taste.


Isn't that how Spotify originally did their discover playlists? Seems like I remember they took the songs you listened to most frequently, found playlists other users have made containing that song, then looked for songs you haven't listened to on the platform yet that frequently appeared in playlists with the songs you play frequently.


Interesting idea!

Slightly off-topic, but originally I had planned to organically grow the playlist count by allowing users to login with their Spotify account and reading their playlists. I even made a prototype...but turns out Spotify only allows working with user data once you have your API quota extended, which seemed like quite the undertaking.


Yer their quote system is a bit painful. I built a bunch of background jobs that continually retrieved data ... max'ed out the quote limit constantly.

I even went as far as setting a second developer account to have a larger limit ;)

I also did a writeup[1] on lessons learnt in case someone is interested in rinsing and repeating on this idea.

[1]=https://gregorius.rippenstein.art/works/hamoni


I'll check out the write-up, thanks. As for user-data, the request count limit is not even the limiting factor. If you want to request data with a users token, you can't at all, without getting a quota extension (apart from 25 pre-registered developer users).


https://tunemeet.com/ has kind of similar idea, it matches you to other people who are listening to the same song at the moment. Or to people who have some of the top 5 songs as you


if you think this is a good idea, just go to music shows for your favourite artists. A lot of people go to those solo. If you talk to someone you like, and you spend a music show together dancing and enjoying an artist you both really like you are going to have such a strong start of the relationship it's only going downhill from that momement [0]

[0] true story, multiple times


spotify constantly calls the facebook SDK to tell it everything you’re doing and listening to, so Facebook has the data although they’ve failed to build a dating app.


stats.fm app has a matching service using spotify stats that sound similar to what you are describing


Nice start - I think there is ultimately some value in this approach, but currently seems quite limited in its functionality, for example I can only choose one song? This results in a very large set of results to wade through, many of which will not be helpful or relevant.

Personally I'd get far more value if I could chuck in 3-4 favourite songs and see which playlists contains all of them (or order by greatest number of matches first); this would allow me to more accurately match my tastes to discover a potentially new bunch of other songs I might like.

If there is a way to match against more songs, apologies! The post title implies there is, but the site seems to suggest it's limited to one song.


Ditto, this is what I thought it was -- the title says "by songS they contain", not "by a song they contain".

I really want to see if anyone else has Left Hand Free and Pop Muzik on a playlist together, and if they've found any _other_ songs that contain eeny-meeny-miney-moe in the lyrics and made a playlist of them like I have...

(Here's another oddball one, if you're a puzzle type. Bring The Magic - Steve Wynn. So It Goes - Nick Lowe. Roadrunner Once - Jonathan Richman. Heavy Metal - Sammy Hagar.)


I have extremely niche playlists like this: male and female presenting singers shouting the same lyrics at the same time, songs with cheerleaders in the background, songs by two or more NYC area rappers about high seas piracy. For your list consider adding "Hot" by Liili in which the lyrics code switch Russian and English but do contain "eeny-meeny-money."


Thank you for the feedback!

It is currently limited to one song, but there is already an issue on github planning multiple song selection. Just haven't gotten around to it yet.

As for over-abundance of results, sorting by name length in ascending order should help at least somewhat, but the search approach is an open problem.

As an aside: on large screens there is also an option to choose a specific song by dragging it from the Spotify client.


brilliant idea. i already found a couple awesome playlists by repurposing google to be this tool

https://www.google.com/search?q=%22state+of+grace%22+%22I%27...


This kind of thing has always seemed quite interesting (and worrying!) to me from an OSINT perspective. It seems like if you just know a couple of some anonymous person's favourite songs from various genres, you could identify possible Spotify accounts they may also own. As far as I can tell, it's basically impossible to discuss anything remotely personal online without jeopardizing pseudonymity.


Most people that know my taste in music already have my spotify because they're my friends and I gave it to them. As for worrying about pseudonymity, spotify lets you hide listening activity and make all your playlists private.


It's kind of horrifying to learn that it only takes 33 bits of information to uniquely identify any living human on the planet. log2(8e9) ~= 32.9.


If it helps (with the horror I mean), that's just a sort of lower bound of course - if you could choose your bits carefully and ideally (and I suppose they'd probably be pretty weird (to a human) and overlapping/multi-dimensional, like 'lives in Europe or Antarctica' or 'uses macOS and is female') which of course you couldn't.

i.e. in practice, for practical metrics, it probably takes a lot (I'm not going to guess how many) more.


What data would be useful from finding someone's Spotify? I dont use it so I'm not sure what details can be compromised.


Because there's a good chance they use the same username on other services. If you didn't already know that, now you might.


Yep, plus I know a lot of my friends using Spotify login using Facebook, which displays their full name & profile photo. Also followers/following lists are public, so if you can find a close contact of theirs who is less serious about opsec, you're that much closer to their actual identity. This is only really relevant to the truly paranoid who try to use a different identity for every service.


I tried it with a few tracks from a very popular playlist ('Yoga Electronica').

First track couldn't be found. Second track could be found, but it couldn't find the aforementioned playlist (with almost 200k likes). Third song also can't seem to be found.

So I'm thinking that the indexing is not complete. Unless I misunderstand something.


Thanks for the feedback!

Currently the database contains only ~18k playlists and ~1.4M tracks. So only the relationships between them are used in the app.

Edit: theres even a counter at the top of the page, but it takes a while to show up, because of the HN effect.


Spotify playlists.


Lol, the first song I searched for game 0 results. (Searched for "More With You").

Oh and a bug: If I first search for something and get a result, then enter a new search and scroll down before this new search has completed, it starts searching for page 2-7 of this new search term. Even better if you've already triggered the search for page 2-7 on that first term. If you then scroll down after updating the search, it is now searching for page 8-11 of that new search term :)


Thanks, I registered the issue on GitHub.


It seems that within 10ish minutes the site has already been hugged to near-death. Thank You :D


Nice idea. I'd recommend adding some kind of visual indicator when it's in a loading state - at first I thought it wasn't working but it was just slow to respond (probably because it's close to the top of HN right now...)


Thanks for the feedback!

The loading indicator is also on the list, TBH was supposed to go up before ShowHN, but I forgot. And yeah, now the query execution wait time for my measly 2-core postgres instance is about 30 seconds :D.


I tried to search for a song I found to be quite cool, Rotoscoping by PTRD. I got two matches, one "Discover Weekly Archive" made by Skiley.net on behalf of NomadicDaggy (you, the creator of this post and service). The second one was a "On Repeat"-playlist! Here, I realized that it's possible to look at other people's "On Repeat"-playlists, even though they are styled by the Spotify app to look like it's your own.

It's kind of weird that I only found this song in your playlists, and not in one of my playlists which also contains this song. For example: https://open.spotify.com/playlist/0IntLgAqI2EysEvqwiivXQ?si=...

Anyway, I'm checking out your "On Repeat" playlist, however it might be better if your personally generated playlists are excluded? I'm not sure.

Really cool service!


Thanks for trying it out! The site only includes indexed playlists and tracks, which have to be updated semi-manually, so currently only 18k playlists can be found. I plan to increase this, of course, but not at a crazy rate, because I'm hosting this out of my own pocket.


For that I regularly use this free service https://www.chosic.com/spotify-playlist-search-tool-by-song-...


This is alternate approach, that uses an advanced Google search under the hood. It works very approximately, given that it searches the song title in everything Google decides to index.


Thank you, was not aware of how it works


I'm not a 100% on how it works myself, just a stong hunch based on the site speed and results.


This feedback is coming from someone that wants to discover new / unknown bands in specific genres. I have a playlist of new / unknown bands I've discovered over the years, but searching with your tool didn't give back anything.

The playlist in case you want to use it for testing later: https://open.spotify.com/playlist/5DcbwcCkBde68RgsxO2Sn2

Ideas for improvement:

* autocomplete

* autocomplete with song, artist and album image

* allow for search by track & artist name

* match up one of my playlists with a playlist you have in your DB


You can generate new songs with Echoes app. Connect your Spotify account, you will explore top artists, top songs and recently played. Also New Discovery section will generate playlists in a specific genre, if you click Load New Tracks, it will generate new songs through your Spotify algorithm.

https://echoesapp.io


Thanks for the feedback, I'll look into it!


Cool project!

I maintain an archive of Spotify playlists, scraped daily via GitHub Actions: https://github.com/mackorone/spotify-playlist-archive.

If you're willing to share, I'd love to hear about your experience with the Spotify API. What endpoints do you use? How often do you call them? How do you handle rate limiting? Has Spotify reached out regarding storing playlist data (being in violation of their terms of service)? Any other lessons or advice?

Additionally, I wonder if there's any opportunity for collaboration here. Perhaps you can use the archive to fill out your dataset? Maybe it can act as a cache for your system's scraping needs? Perhaps our Web UIs could link to each other or be combined somehow (https://spotifyplaylistarchive.com/)?

> Currently we keep track of 18378 Spotify playlists

How did you find all of these playlists? Querying Spotify for "Spotify"-owned playlists gives me less than 5000 results. I'd love to add more playlists to my archive but I don't have good discovery mechanisms, especially for non-Spotify-owned playlists.


This is neat! I built a similar service that basically uses song playlist co-occurrence to generate song recommendations for any given Spotify track, and have found some good initial success for more popular tracks / artists [1].

The bottleneck here is using the Spotify API to find relevant playlists - as others have noted here, Spotify's API doesn't provide a way to perform direct lookup of playlists containing a given track, so the best approach I've found is to perform many text-based searches for desired search terms (e.g. artist, song name), and then do breadth-first-search on playlists created by the same users from the initial result set, in order to find other playlists that might have the artist or song you are looking for.

[1] Available in very limited fashion at https://vybe.link; this is a Spotify beta application so the full app only works for whitelisted Spotify accounts


Cool concept. The site seems to be succumbing to the HN effect. There should be some kind of loading indicator on the first search, it didn't seem to be doing anything but when I went to close out the page there were results. I typed "One Beer" into search and a lot of duplicate entries came up.


Thanks for the feedback! The loading indicator is an overlook on my side and you are not the only one that noticed :D. I seem to have been very naive about how HN might affect the site.


Thanks god, been waiting for this concept to be implemented in a better way


Thanks for this! I tried 3 songs and they came back not found. But on the 4th, I got a good hit! (Mix of Progressive, Deep House & Trance, designed for High Intensity Focus & Created for my Medical School Studying by Austin Nguyen)

https://playlists.dags.dev/tracks/7xdkou0YaNLW9hWXU9MRcR/pla...

Tried a bunch more without success. Granted, I like a lot of obscure electronica.


Thanks for trying it out! There currently are only 18k playlists indexed, so only those can be found. I've been planing a feature, where users can suggest playlist search terms by which I should extend the index.


When on dark mode the results are really card to read fyi.

Love this concept, way quicker than using the “found on” in the Spotify app - which doesn’t even always work that well.


Thanks for the feedback, I'll look into it.


I have wished for something like this, so I tried it out, using _Beauty Beats_, by Beats Antique. I got one of your own playlists - what are the odds?!

Works great, thank you!


Awesome, thanks!


Love it. This is one of the primary features that RDIO had that I miss in Spotify, because it allows for discovering songs, as in “if someone likes this song enough to add it to a song list, I wonder what else they like”. I look forward to the indexing to increase, because many of my favorites aren’t in your index yet. (And they what would it be like if you cooold enter two or more song titles?)


I think searching by the spotify song id / song link, would be easier in my opinion, as I couldn't really find what i was looking for.


That functionality already exists, but is prompted for only on large screens, because I didn't think anybody would use it on mobile. And thanks for the feedback!


Thanks, checked it out on my laptop. Great stuff!


This is very convenient, but it didn't work well for me on the first attempt. The song I used is called "Love", from Colour Haze.

> Found 32646 tracks.

Yikes. Adding the artist to the search yields 0 results, and since it only loads a number of results until you scroll to the bottom of the page, I could not find the track I wanted.

It would be great to have the option to specify the author somehow


Spotify already has a "Discovered On" section on the Artist page which has playlists that include the artists.


I’ve maintained a playlist for years of my favorite sad songs.

Mostly acoustic and piano based, lesser-known songs. Please add it to your index.

https://youtube.com/playlist?list=PLhMrEC5L8T4FmbQP78B5gzEJh...


Currently the site only works with Spotify tracks and playlists. If you care to create a public spotify playlist, I'll gladly add it in.


Found some playlists for a few songs. But doesn't work for others. Says Spotify track link should look like 'https://open.spotify.com/track/22numbersOrLetters'


Spotify API is pretty cool. I created https://hotkeyplayer.com/ with it to quickly map keys to playlists and launch them with the press of a button. Used for switching music during D&D sessions


This is very helpful! I've searched for this in the past. There's sometimes one song I like and I'd like to find more like that.

Something useful I found is the "Go to song radio" in Spotify, that creates kind of a "dynamic" playlist with similar songs.


Thank you very much!

I also use the Spotify radios a lot, but they are heavily curated - they contain a lot of songs that you already have in a playlist or have liked. So they are useful for listening, but slightly less for strictly music discovery.


I use radios in a hybrid approach. I'll make a radio, find a song I like, and then go play the album through and let Spotify start recommending, and repeat. Tends to make enough variety to get me to all sorts of interesting places


Amazing! Something I've been meaning to build myself and now I don't have to. I'd like to echo the need for multiple song search—my use case for this was good playlists in cafes where I want to identify the playlist based on 1-3 shazam'd songs.


Thanks and noted! :)


I need something that can find the song from my out of tune, only partially remembered humming


Melodic contours might help you[0]--you only need to remember if the tune goes up or down.

[0]: https://www.musipedia.org/melodic_contour.html


https://www.reddit.com/r/NameThatSong/ might be able to help with that


SoundHound used to do that, looks like it can still be used through: https://www.midomi.com


Why "used to"? The SoundHound (Android) app is still around...


I just briefly looked at their website and saw the B2B pivot. Wasn't sure if the app is still around, but upon closer inspection it seems like it is.


I use 1001 tracklists for this - you can search by track and it will surface DJ sets where the song was played.

https://www.1001tracklists.com/


I made a site to record and vote on playlists sorted by last updated as well as popularity https://playlists.mohiohio.com


You can also use https://volt.fm for this. And it currently has 337,502 playlists indexed.

Disclaimer: I've built this.


Nice timing. I need a good danceable rock playlist by the end of the week. I could create my own, or I could use some danceable rock songs to find an existing playlist.


Great to hear! Just not that the site is very unresponsive right now, thanks to all the HN love. So probably a good idea to bookmark it and come back later.


Can someone please build an inverse recommendation system? These recommender systems are too good at giving me things similar to what I've already heard.


I tried a song called "Chase Init_" by dragging and dropping but it doesn't work. A Wingdings-song isn't a problem though for some reason.


How does this work on mobile? There’s no submit form button


Thanks for the feedback, I'll add a button. For me it works just by pressing "enter" on the mobile keyboard, but admittedly I didn't test it very thoroughly.


Maybe I tested it during hug of death. Return “worked” (I could press it) but there was no indication given even by the browser progress bar that anything was happening as a result


Apple Music already has this feature. Does Spotify not?


Other people have made similar tools, but Spotify still hasn't added any native playlist-search-by-tracks AFAIK.


Could you point me to some detail on how to actually do this in AM?


If you search on Apple Music (and actually tap “search”, don’t use the autocomplete results), there will be a list of categories on top. Playlist is one of those categories. You need to type out song name and artist when you search, otherwise it will just search for playlists by that name.

Obviously this could be a lot better.


Got excited. Saw it was Spotify. No longer excited.


This is really nice. I especially like the feature that let's your drag a song from a UI into the app. Very handy.

Vibin to a new playlist now.


This is great, I popped in a song title from a band I like and found a playlist with some tracks that sound really good!


I always thought something like this for movies that contain a favorite song might reveal a hidden gem or two.


It's kind of mind boggling to me that Spotify doesn't offer this themselves. Seems like use case 101


There's a post in their official forums with 60+ pages from 2012, but they state that there's not enough demand...


Like literally everything in their forums. There's a reason I dropped them.


You have spotify radio which is very similar.


Thank you so much for this! I've always wanted this capability, and I plan on making extensive use of it.


Awesome, glad you like it!


I'd love to see a playlist of submitted songs that aren't part of your playlist scrape.


Good start, but I'm not sure how song title can extract meaningful similar songs. I put in "Big Time" (song by Peter Gabriel) and it came back with a list of songs from a variety of genres. From smooth jazz selections to songs by Bjork. I think you really need a machine learning approach that categorizes songs by their signal content into categories.


Thanks for the feedback! You are supposed to click on a song to see playlists it contains. Sorry for the unclear UI.


ahh I see. Yeah that makes more sense.


What do I have to do after entering a song? Pressing Enter does nothing (Firefox, Safari, MacOS).


As other people have noted, the site is missing a loading indicator and currently, very many people are trying to access it, so you are doing everything correctly, just currently requests take upwards of 30 seconds to process.


thanks


Thanks, I was wondering what I'm doing incorrectly (Firefox, Linux).


I'd love to see a playlist of songs you don't have in any playlist.


Nice idea. Out of curiosity, which song inspired you to build this site?


To be honest I can't remember anymore, it was a few years back that I initially got the idea for it. Might have been Alone by The Upbeats.


“Boobs” by Ruth Wallis doesn’t appear to be indexed by this site?


it doesn't work for me. that, or I feed it obscure songs.


Spotify has had a pretty awesome public api for a while


It looks great on paper but the quotas and lack of support make it painful to use outside of toys. The dev forums are full of these complaints, and before that the github repos were too. We are long overdue being moved to another platform so they can more legitimately ignore the same old support issues and requests they don't intend to solve. Websockets is a good example, libspotify successor another. Frustrating.


Spotify sorely needs this feature. Great work!


Cool!

I Would love something like this for Youtube.


thank you




The deadline for YC's W25 batch is 8pm PT tonight. Go for it!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: