What did I build?
This is Kipify, a web app that helps you find new music on Spotify. Kipify analyses your Spotify data from the past to determine what you enjoy and what you don't, and then accurately (it tries its best) generate recommendations.
What makes it different from Spotify recommendations is that you can also get tracks, similar to specific tracks. To use the app all you have to do is login with your Spotify account, and hit create on one of the two methods provided. Kipify will then automatically make a playlist of 50 tracks.
Why did I build it?
It originally started as a final project for Harvard CS50 Final Project, but then I decided to take it much further considering I myself need something like this. Kipify isn't a paid service, it's actually open sourced. Here's the GitHub repository with all the source code: https://github.com/kipifyapp/webapp
Although the project was completed in October 2023 itself, I'm only posting about it now because I had school and couldn't really find the time and I forgot.
How did I built it?
No I did not design the logo, Microsoft Bing AI did it for me. Yes I came up with the name.
Now here is the technological part you have been waiting for...
Kipify was built using Node.js, Tailwind CSS, Express.js, and the Spotify Web Api. It is hosted with the help of Vercel.
I was initially going to use Next.js, but then it turned out to be a little too complicated for me given how much time I had. So I chose to use plain HTML with Tailwind css.
A little command helped me make sure the site looked pretty:
"scripts": {
"tailwind": "npx tailwindcss -o public/styles/build.css --minify"
}
Yes I used npm run tailwind every time I made a change to the styling. Not at all the best way to develop, but fair enough...
Although I had to little extra bits to add to the backend, I still chose Vercel because it has much better features, it's free while still being performant (serverless functions).
The "Algorithm"
The user is provided with two ways to create a playlist
Create from profileorCreate from tracks.In case the user selects the first option, the top 5 tracks they have listened to the most in the past month are selected. We get this data from Get User's Top Items API.
In case the user selects the second option, they have to provide URLs for 1-5 Spotify tracks.
After that we get the audio features of all the tracks using Get Several Tracks' Audio Features API.
Then one by one we get the recommendations for each track using Get Recommendations API.
All the recommended tracks are ranked according to how many recommendations they appear in. And then the top 50 tracks are chosen from that. Let me give you an example:
Track 1 Suggestions: A, B, C, G
Track 2 Suggestions: B, D, F, H
Track 3 Suggestions: A, B, E, I
Suggested tracks list:
- 3 (appeared 3 times): B
- 2 (appeared 2 times): A
- 1 (appeared once): C, E, F, G, H, I
If only top 5 tracks are to be selected,
the final generated tracks list would be:
B, A, and 3 more random tracks from `1`
The 50 generated tracks are then put into a playlist using Add Items to Playlist API.
The playlist name is put as
Kipify Mix, the reason is that Spotify creates personalized playlists a lot of times, likeRock Mix,Pop Mix, etc. Therefore the playlist name containingMixgives the impression that it is a playlist created by Spotify itself, which would attract anyone stalking our user's profile. The playlist description is put asPlaylist created by kipify.vercel.app for <username>. Yet another little trick to get more people to know that Kipify exists.
A lot of people I have told about this app have had good experience so far. The generated playlist is fairly accurate and definitely helps in finding new music.
I do aspire to build this app bigger and better in the near future (probably not in the next 2 years), going beyond recommendations. Maybe a way to automate the whole process of migrating all your playlists to YouTube Music.....
Here's the link to the web app incase you missed it above: https://kipify.vercel.app/
Looking forward to the feedback in the comments below :D
p.s ive been using spotify since 4 years and i still have not paid for premium
Top comments (22)
Cool app! Here are some things you could consider:
By default when clicking on the create button on the home page, it checks for login, but not on the create button in the nav bar, I see what needs to be improved.
Definitely a very good idea to make the whole experience simpler.
That sounds interesting, however there's also a slight chance that one track on a platform is not available on another platform.
Oh and if you you reverted your algorithm, you could use it as a "discovery" playlist instead of "similar songs", i.e. the songs that got recommended by the API the least will be selected :-)
Hmm... I didn't think about this.
It is scary how well this worked. All songs I never heard before but were relevant to my tastes. Great job!
Thanks, glad you enjoyed.
Awesome, thank you very much
Your welcome!
I enjoyed reading this post, especially the algorithm part. Good job!
Thank you!
How do you plan to make it better?
That's a question I can't exactly answer right now. But a rough idea would be improving the algorithm and the overall user experience, making the app available on IOS and Android.
Pretty sweet. Always open to new music. Thanks! Best of luck building the app out.
Have fun listening to songs
Great work here, something I personally need. Will give it a spin and share feedback! Itโs cool that itโs open sourced too
Glad I could help out
Nice work on Kipify. Interesting use of technology and it seems like a useful app. Good luck with your future projects
Thank you
Great app and great post! I didn't know what you can do with Spotify for Devs.
Thanks
Some comments may only be visible to logged-in visitors. Sign in to view all comments.