Jump to content
Search In
  • More options...
Find results that contain...
Find results in...

Phosphorescence, a Spotify-driven trance music playlist builder

Recommended Posts






Phosphorescence is trance music playlist builder powered by the Spotify API. It is created by me and this is the beta launch :) Playlists are cohesive in track selection both in the sense of harmonics and BPM, as well as in the sense of "evocativeness". It has been built in a very pluggable way as I have plans to make it even more exciting to use for people who wish to script their own playlist building within my framework.


The best experience is using Chrome with Spotify Premium. Opera, Firefox, and Edge should all work just as well (although Edge I can only test through the VM so please let me know if it's not working). Safari does not support in-browser playback but falls back to a simpler form of the application gracefully. IE is not supported. Spotify free users will not be able to control playback in their browser at all; they can generate playlists and save them to their account.


This is in beta. It's possible (hopefully not) that your first impression is awful due to some unforseen bug. Please let me know here if that happens, preferably with system info and a screenshot and some description of events leading up to it. I've been personally using the product for a while now and fixing bugs but it's hard to test all user behavior when it's just yourself set in your ways.


Here is the link:




I know Doomworld is more metal than trance but I seem to remember some level of trance music contingent here.




I've been putting off creating a thread for this for much longer than I originally thought I would.


Back in January I had this idea on how to classify trance music which is one of my favorite genres. This classification didn't go into the nitpicky subgenre rabbit-hole that is common with many genres of music but rather was made up of an idea of two-axis "compass" where tracks could fall. Initially just a fun idea for armchair classification, I quickly realized there must be a way to actually do something cool with it and test it out.


I've been using Spotify for ages and it turns out they acquired an audio analysis company some time back and now make available through their API a ton of analyzed audio data all for free (assuming you follow the developer TOS).


So I began diving into machine learning, something I was aware of and knew the basic ideas of, but had never ventured into. I can say that what I came up with is very simple, using existing libraries for regression analysis, but I excited by how well the training seemed to be working.


Once the ML model was built, I started work on the actual application and after tons of iteration and personal use-driven bug fixing and enhancements and some month-long breaks I finally made the push to release this into "beta" (basically I tell people who I don't personally know about the website).


There's a lot in the pipeline as well, depending on how things play out. The entire track building system is fully pluggable and I have some preliminary stuff to make writing your own set-building scripts easy. However for those who don't know or care what that means, I have made the basic user experience very approachable from a one-click and go perspective.


I won't bore people with the rest of the story, basically here it is, in beta. One of the few side projects I actually ended up making public.


Share this post

Link to post

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now