• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Amirs' test tracks (Spotify playlist + how AI "hears" them)

pavuol

Major Contributor
Forum Donor
Joined
Oct 2, 2019
Messages
1,562
Likes
3,950
Location
EU next to warzone :.(
I wanted to share some interesting info about the data hidden behind songs that streaming services use for their analytics and AI purpose (later on that topic).

To be more beneficial for ASR readers (or as a side effect) I decided to make a playlist of songs Amir mentioned in his reviews (consistent sound quality in one place).

Notes about playlist:
Not meant to be exhaustive nor official, I have included just songs Amir explicitly lists within main reviews. Sometimes he mentioned only the whole album and no particular tracks so I omited that cases (I can add some after further specification :) )

link to playlist

Gait Kelin Kromhof – Gait
Eleni Karaindrou - Medea
Roger Waters – Amused to death (broken YT link to specific track)
Nils Lofgren – Acoustic live
Kings’s singers – Good vibrations
Samuel Orson – Cascadia

Polk T50 Equalization EQ FLOOR STANDING TOWER SPEAKERS.png

Roon.PNG


Now to the "raw numbers" behind songs. Lately I searched if there is some tool for saving Spotify playlists as text/spreadsheet files. I found Exportify which manages the job perfectly. What was interesting, besides the usual tags it also extracted some metadata I believe Spotify AI engine uses for purposes such as automatic song/album/playlist suggestions, auto-generated playlists, user behaviour analytics etc. This data I think are just part of the more complex story. This makes me wonder just how complicated contemporary AI engines of leading streaming services can be and how many algorithms they use...

List of interesting columns with raw numbers extracted from tracks:
Danceability /Energy / Key / Loudness / Mode / Speechiness / Acousticness / Instrumentalness / Liveness / Valence / Tempo

You can download full spreadsheet (csv / xls) or you can export your own playlists and have "fun" seeing what track is the "most danceable", the "loudest", bears "the most energy" etc. :) Or if you are active Spotify user you can use Exportify for some crazy complex analysis of your music taste like the author of Exportify did with his playlist (Pavel Komarov).

/ edit: added all tracks hinted in post #728
 
Last edited:

NoFuzz

Member
Joined
Sep 11, 2019
Messages
15
Likes
14
Great list, thanks! It's what I tend to like best about any equipment review: the music to discover!

I've converted it for Tidal users:

https://listen.tidal.com/playlist/e277c1ea-e881-469a-aa7f-4494a281e653
Updated:

 

Hark01

Member
Joined
Feb 7, 2021
Messages
22
Likes
14
Top Bottom