3 ms·

With the luxury of working from home during the pandemic, I've much preferred using the opportunity to discover new music during the day, rather than listening

With the luxury of working from home during the pandemic, I've much preferred using the opportunity to discover new music during the day, rather than listening to nondescript ambient music on the loop. I suggest others give it a try. It's not like the two are mutually exclusive. You could give Robert Fripp's discography a try for ambient/electronic, for instance: https://open.spotify.com/album/3gHpX5wJn0kzAXJvwORG6u https://open.spotify.com/album/3gHpX5wJn0kzAXJvwORG6u

throwaway675309 5y ago
It depends on the type of work I'm doing, for me the issue with new music is that the novelty makes it so engaging that it may distract me from being able to fully concentrate on my work which is why sometimes I definitely prefer a more ambient chill vibe. Same thing happens to me with music of sufficient complexity such as a baroque counterpoint style.
- airstrike 5y ago
  I'm sort of the opposite... I need absolutely new music to be able to work with something on the background. Otherwise I recognize the song and start humming along...
kristopolous 5y ago
I've been working on a music discovery platform for, well, really 14 years but my latest rewrite I started about 18 months ago now. It's not approachable yet, but based on the response here that may change. It is primarily label based. By using discogs, essentially imdb for music, it uses some Foaf and affinity "distances". Here's an example flow, let's say you find an artist you like, say Cygnus (https://www.discogs.com/artist/1932807-Cygnus-5 https://www.discogs.com/artist/1932807-Cygnus-5). We pull down the releases and get the labels. Then go to the labels which have links to things like Bandcamp, SoundCloud, or various other sites. The tool then catalogs the releases and adds them essentially to an unlabeled dataset. (I've been updating dead links on discogs for a while to make this thing with better btw ;p ) Then there's another tool that goes through to "label" things by you listening to them. Essentially it's designed to do a single pass. There's a REPL and it interfaces with mpv using their JSON and Lua interface. Then you rate each release and the adjacent information, such as the people working on it (on its discogs catalog page) becomes another vector for possible traversal. It's not a constant scraper, it'd go on forever. Essentially it's manually run and there's a tweakable threshold. For instance. Say you highly rate multiple releases that share a producer, artist, and label. Well, by golly, if there's a few other releases in that grouping you'll probably like that too. I mean, duh, obvious, right? A lot of this approach came from the world of italo disco where the real names often got obscured behind short lived projects and labels which required a bit of digging to find the related content. Take Aldo Martinelli for instance. Here's his fb banner which is the best catalog of his work I know (https://scontent-lax3-1.xx.fbcdn.net/v/t31.18172-8/fr/cp0/e15/q65/882303_423096714449569_493670927_o.jpg?_nc_cat=110&ccb=1-3&_nc_sid=dd9801&efg=eyJpIjoidCJ9&_nc_ohc=xtSpMTmYBoQAX98GsEQ&_nc_ht=scontent-lax3-1.xx&oh=be2add42c01349c2d157bc56ab174522&oe=611CE274 https://scontent-lax3-1.xx.fbcdn.net/v/t31.18172-8/fr/cp0/e1...). Same person doing all of that. For instance, "topo & roby" and "moon ray comanchero" are essentially the same people, but getting from one to the other or uncovering the relationships, that's the challenge this infrastructure tries to solve. This stuff gets unusually complicated with specific regards to italo disco. See the profile of radiorama for instance (https://www.discogs.com/artist/73973-Radiorama https://www.discogs.com/artist/73973-Radiorama) or groups like The Flirts, Fun Fun or Den Harrow, where models were the faces, singers were the vocals, different song writers and musicians would come and go, they were all huge tangles. This method completely ignores popularity or what others are listening to and falls back on the networks of the people making the content. For me at least, it's produced great results. But it's built exclusively for me at this time. I don't even have most of the traversal code posted The code is over here https://github.com/kristopolous/music-explorer https://github.com/kristopolous/music-explorer I'd be happy to make it more accessible but I'm multiple standard deviations from the mean when it comes to how I approach music so I've got no idea whether there's a large interest here If you want to see what results look like, now this is exclusively catered to my tastes, I have a pretty basic hack to listen to things on the go over here http://9ol.es/pl/ http://9ol.es/pl/ I guess the focus here is to see the label/artist diversity even if the style of music isn't your thing. --- As far as listening while you work, the tool is instrumentable. For me I can press a quasimode key (say the windows key) along with some other key and do things like pause/play, label the release, track/release navigation, get info, open the webpage that it was on, etc. These are hooked at the window manager level. This way there's as little task switching as possible and I can stay in whatever application I'm doing.

throwaway675309 5y ago

It depends on the type of work I'm doing, for me the issue with new music is that the novelty makes it so engaging that it may distract me from being able to fully concentrate on my work which is why sometimes I definitely prefer a more ambient chill vibe. Same thing happens to me with music of sufficient complexity such as a baroque counterpoint style.

airstrike 5y ago
I'm sort of the opposite... I need absolutely new music to be able to work with something on the background. Otherwise I recognize the song and start humming along...

kristopolous 5y ago

I've been working on a music discovery platform for, well, really 14 years but my latest rewrite I started about 18 months ago now. It's not approachable yet, but based on the response here that may change. It is primarily label based. By using discogs, essentially imdb for music, it uses some Foaf and affinity "distances". Here's an example flow, let's say you find an artist you like, say Cygnus (https://www.discogs.com/artist/1932807-Cygnus-5 https://www.discogs.com/artist/1932807-Cygnus-5). We pull down the releases and get the labels. Then go to the labels which have links to things like Bandcamp, SoundCloud, or various other sites. The tool then catalogs the releases and adds them essentially to an unlabeled dataset. (I've been updating dead links on discogs for a while to make this thing with better btw ;p ) Then there's another tool that goes through to "label" things by you listening to them. Essentially it's designed to do a single pass. There's a REPL and it interfaces with mpv using their JSON and Lua interface. Then you rate each release and the adjacent information, such as the people working on it (on its discogs catalog page) becomes another vector for possible traversal. It's not a constant scraper, it'd go on forever. Essentially it's manually run and there's a tweakable threshold. For instance. Say you highly rate multiple releases that share a producer, artist, and label. Well, by golly, if there's a few other releases in that grouping you'll probably like that too. I mean, duh, obvious, right? A lot of this approach came from the world of italo disco where the real names often got obscured behind short lived projects and labels which required a bit of digging to find the related content. Take Aldo Martinelli for instance. Here's his fb banner which is the best catalog of his work I know (https://scontent-lax3-1.xx.fbcdn.net/v/t31.18172-8/fr/cp0/e15/q65/882303_423096714449569_493670927_o.jpg?_nc_cat=110&ccb=1-3&_nc_sid=dd9801&efg=eyJpIjoidCJ9&_nc_ohc=xtSpMTmYBoQAX98GsEQ&_nc_ht=scontent-lax3-1.xx&oh=be2add42c01349c2d157bc56ab174522&oe=611CE274 https://scontent-lax3-1.xx.fbcdn.net/v/t31.18172-8/fr/cp0/e1...). Same person doing all of that. For instance, "topo & roby" and "moon ray comanchero" are essentially the same people, but getting from one to the other or uncovering the relationships, that's the challenge this infrastructure tries to solve. This stuff gets unusually complicated with specific regards to italo disco. See the profile of radiorama for instance (https://www.discogs.com/artist/73973-Radiorama https://www.discogs.com/artist/73973-Radiorama) or groups like The Flirts, Fun Fun or Den Harrow, where models were the faces, singers were the vocals, different song writers and musicians would come and go, they were all huge tangles. This method completely ignores popularity or what others are listening to and falls back on the networks of the people making the content. For me at least, it's produced great results. But it's built exclusively for me at this time. I don't even have most of the traversal code posted The code is over here https://github.com/kristopolous/music-explorer https://github.com/kristopolous/music-explorer I'd be happy to make it more accessible but I'm multiple standard deviations from the mean when it comes to how I approach music so I've got no idea whether there's a large interest here If you want to see what results look like, now this is exclusively catered to my tastes, I have a pretty basic hack to listen to things on the go over here http://9ol.es/pl/ http://9ol.es/pl/ I guess the focus here is to see the label/artist diversity even if the style of music isn't your thing. --- As far as listening while you work, the tool is instrumentable. For me I can press a quasimode key (say the windows key) along with some other key and do things like pause/play, label the release, track/release navigation, get info, open the webpage that it was on, etc. These are hooked at the window manager level. This way there's as little task switching as possible and I can stay in whatever application I'm doing.