Slow Build x 10^6
September 19, 2011 2:37 PM Subscribe
By processing a million songs in twenty minutes, and using the Stairway detector Paul discovered many songs that Slow Build "more" (up to 29) than Stairway to Heaven (which gets only a 9). [via]
Paul Lamere is also responsible for Six Degrees of Black Sabbath, previously. He's analyzed the Loudness War in music production, covered on the blue: louder, over. He helped us find Swing and his company brought us the tools that Mefi's own cortex used for a couple of projects.
He was inspired by posts on Reddit asking for slow builds. Yep, we've got those too, even in film.
Previously on data mining movie metadata.
Paul Lamere is also responsible for Six Degrees of Black Sabbath, previously. He's analyzed the Loudness War in music production, covered on the blue: louder, over. He helped us find Swing and his company brought us the tools that Mefi's own cortex used for a couple of projects.
He was inspired by posts on Reddit asking for slow builds. Yep, we've got those too, even in film.
Previously on data mining movie metadata.
Yeah, is there no way to search the songs? I'm curious to know if King Crimson's Starless is in there, but can't be bothered to scroll all the way through all of those images.
posted by saladin at 2:49 PM on September 19, 2011
posted by saladin at 2:49 PM on September 19, 2011
This is one of the best metafilter posts ever. Good job!
posted by Kwine at 2:53 PM on September 19, 2011
posted by Kwine at 2:53 PM on September 19, 2011
With a score of 0.12, Bad Romance has no slow build at all.
posted by swift at 3:01 PM on September 19, 2011
posted by swift at 3:01 PM on September 19, 2011
> is there no way to search the songs?
Forgot to link the web app that lets you search the database. The discovered link above is the key one that talks about the process and links to the web app.
Some of the track metadata is screwed up. David Bowie's Heroes is listed with a title of "David Bowie - Heroes" (artist concatenated into title). It gets a 1.66, but the neat thing is the graph. Maybe another metric for builds even if they aren't all-track-long would help?
posted by morganw at 3:02 PM on September 19, 2011 [1 favorite]
Forgot to link the web app that lets you search the database. The discovered link above is the key one that talks about the process and links to the web app.
Some of the track metadata is screwed up. David Bowie's Heroes is listed with a title of "David Bowie - Heroes" (artist concatenated into title). It gets a 1.66, but the neat thing is the graph. Maybe another metric for builds even if they aren't all-track-long would help?
posted by morganw at 3:02 PM on September 19, 2011 [1 favorite]
Adagio for Strings also has no slow build. Most recordings of Khachaturian's Adagio from Gayaneh have negative slow build. So I have to take some issue with his methodology, but I don't have the knowledge or vocabulary to do it, other than to say "these two works that gradually increase in intensity are not recognized by your algorithm."
posted by infinitewindow at 3:09 PM on September 19, 2011 [1 favorite]
posted by infinitewindow at 3:09 PM on September 19, 2011 [1 favorite]
With a score of 0.08, Gimmie Gimmie Gimmie has no slow build at all.
posted by Slack-a-gogo at 3:09 PM on September 19, 2011 [1 favorite]
posted by Slack-a-gogo at 3:09 PM on September 19, 2011 [1 favorite]
I think there must be something more to the slow build than this algorithm picks up if Starless gets a score of 6.05 and Radiohead: Exit Music (for a Film) gets a 4.65.
posted by FreedomTickler at 3:11 PM on September 19, 2011
posted by FreedomTickler at 3:11 PM on September 19, 2011
This is the only list on the entire internet that has Linkin Park adjacent to the Vienna Philharmonic Orchestra.
posted by Wolfdog at 3:11 PM on September 19, 2011 [1 favorite]
posted by Wolfdog at 3:11 PM on September 19, 2011 [1 favorite]
I'm not a musician, although I do play one in my bedroom, but I don't base a song's slow build on decibels alone.
posted by perhapses at 3:12 PM on September 19, 2011 [3 favorites]
posted by perhapses at 3:12 PM on September 19, 2011 [3 favorites]
This raga by Nikhil Banerjee is a prime example of a song with a slow build that is not reflected in the decibel-biased methodology. I shall take my complaints the the authorities and settle this matter once and for all!
posted by perhapses at 3:21 PM on September 19, 2011 [2 favorites]
posted by perhapses at 3:21 PM on September 19, 2011 [2 favorites]
Starless was the first thing I looked for too. Probably too slow a build for the detector.
posted by Casimir at 3:22 PM on September 19, 2011
posted by Casimir at 3:22 PM on September 19, 2011
The Ramones not only get a -1.67, but also a perfect 0.00.
posted by TedW at 3:25 PM on September 19, 2011 [1 favorite]
posted by TedW at 3:25 PM on September 19, 2011 [1 favorite]
As soon as I saw this, I knew Muse would feature prominently. I'd bet if you averaged the stairway build for all songs by a band/artist, they'd come in number 1.
posted by John Kenneth Fisher at 3:29 PM on September 19, 2011
posted by John Kenneth Fisher at 3:29 PM on September 19, 2011
"With a score of 0.00, And Then There Was Silence has no slow build at all."
I would say not true, but you can't argue with numbers!
posted by adamdschneider at 3:31 PM on September 19, 2011
I would say not true, but you can't argue with numbers!
posted by adamdschneider at 3:31 PM on September 19, 2011
Thanks Cortex for cluing Paul and I in on the post. The great thing about Paul's recent experiments is if you think slow builds can be detected through another feature like tempo or timbre, you can easily run your own map reduce job on the freely available million song data set. Amazon even lets you mount it for free as a public data set. Echo Nest catalogs quite a ton of music features per song -- we know the start of each bar and beat and the pitch and sound of each note.
posted by brianwhitman at 3:32 PM on September 19, 2011 [3 favorites]
posted by brianwhitman at 3:32 PM on September 19, 2011 [3 favorites]
I'd bet if you averaged the stairway build for all songs by a band/artist, they'd come in number 1.
I don't know - Godspeed You! Black Emperor or Explosions in the Sky are stairway-builders extraordinaire, although GYBE frequently jumps off the top of the stair to climb a different one within the same "song".
posted by LionIndex at 3:33 PM on September 19, 2011 [2 favorites]
I don't know - Godspeed You! Black Emperor or Explosions in the Sky are stairway-builders extraordinaire, although GYBE frequently jumps off the top of the stair to climb a different one within the same "song".
posted by LionIndex at 3:33 PM on September 19, 2011 [2 favorites]
Shenanigans!! Free Bird has hardly any slow build at all? That was the only song that any of the girls in my school would slow dance to with me. You know, 'cause it turned into a fast dance song about half way through it. I got so ripped off!
posted by NoMich at 3:35 PM on September 19, 2011
posted by NoMich at 3:35 PM on September 19, 2011
Yeah, weird that none of those famous post-rock bands rank (Explosions in the Sky, Mogwai, etc.) An entire genre of music practically defined by trying to create the most epic slow builds ever.
posted by naju at 3:38 PM on September 19, 2011
posted by naju at 3:38 PM on September 19, 2011
Neat idea!
I can see a couple areas for improvement however. Right now the score is just the slope of the line fitting the curve; a separate "epicness" score would be also weighted by duration -- some of the highest scoring songs are just a minute or two long.
Also I wonder if there could be a different criteria for loudness that takes frequencies and perhaps track density into account. As it stands now, songs produced with heavy compression will not score highly (example) even if they musically represent a slow build.
posted by RobotVoodooPower at 3:38 PM on September 19, 2011 [1 favorite]
I can see a couple areas for improvement however. Right now the score is just the slope of the line fitting the curve; a separate "epicness" score would be also weighted by duration -- some of the highest scoring songs are just a minute or two long.
Also I wonder if there could be a different criteria for loudness that takes frequencies and perhaps track density into account. As it stands now, songs produced with heavy compression will not score highly (example) even if they musically represent a slow build.
posted by RobotVoodooPower at 3:38 PM on September 19, 2011 [1 favorite]
I don't know - Godspeed You! Black Emperor or Explosions in the Sky are stairway-builders extraordinaire, although GYBE frequently jumps off the top of the stair to climb a different one within the same "song".
In fact, the first track on Godspeed You! Black Emperor's Lift Yr Skinny Fists Like Antennas to Heaven has a score of 0.01, which amuses me greatly.
But, you know, there are several links in the FPP that explain, in great detail, how the algorithm works and why the results aren't always what you expect, with specific examples. It's fun to look for cases where it doesn't work so well, but the guy is very aware of the limitations of his approach.
The really noteworthy thing here, in my opinion, is how easy it was for him to do this. A million songs analyzed in 20 minutes! That's amazing! And he basically just pulled together a bunch of stuff lying to hand -- a typical MP3 collection, a few commonplace tools like S3 and MapReduce. This is good work, but none of the pieces are difficult or arcane. I love the exhaustiveness of the presentation, too. I'm going to learn a lot from this once I have a chance to sit down and look really closely at what he was doing. Cool post, morganw.
posted by twirlip at 3:40 PM on September 19, 2011
In fact, the first track on Godspeed You! Black Emperor's Lift Yr Skinny Fists Like Antennas to Heaven has a score of 0.01, which amuses me greatly.
But, you know, there are several links in the FPP that explain, in great detail, how the algorithm works and why the results aren't always what you expect, with specific examples. It's fun to look for cases where it doesn't work so well, but the guy is very aware of the limitations of his approach.
The really noteworthy thing here, in my opinion, is how easy it was for him to do this. A million songs analyzed in 20 minutes! That's amazing! And he basically just pulled together a bunch of stuff lying to hand -- a typical MP3 collection, a few commonplace tools like S3 and MapReduce. This is good work, but none of the pieces are difficult or arcane. I love the exhaustiveness of the presentation, too. I'm going to learn a lot from this once I have a chance to sit down and look really closely at what he was doing. Cool post, morganw.
posted by twirlip at 3:40 PM on September 19, 2011
Oh, hi again Brian! I forgot you had an account when I poked Paul. Heh.
The great thing about Paul's recent experiments is if you think slow builds can be detected through another feature like tempo or timbre, you can easily run your own map reduce job on the freely available million song data set.
Yeah, this is kind of the nutso cool thing that really got me excited about the Echo Nest Remix stuff a couple years back: the fact that you can monkey around with skeletal features of a song at various levels from a measure down to essentially a single transient in the sound file.
The down side is that it's an analytical process trying to glean a whole lot of information out of a necessarily huge variety of recordings on an automated basis so it's neither perfect nor, in some ways, always particularly intuitive to work with at at least the lowest levels of analysis.
The up side is that it's all pretty reasonable to figure out with a little bit of time and effort, and holy shit it's such an amazing toolset. To actually be able to with a really simple script say hey this song should just have one chord and have it (with varying results) just totally work is crazy stunning stuff.
posted by cortex at 3:40 PM on September 19, 2011 [2 favorites]
The great thing about Paul's recent experiments is if you think slow builds can be detected through another feature like tempo or timbre, you can easily run your own map reduce job on the freely available million song data set.
Yeah, this is kind of the nutso cool thing that really got me excited about the Echo Nest Remix stuff a couple years back: the fact that you can monkey around with skeletal features of a song at various levels from a measure down to essentially a single transient in the sound file.
The down side is that it's an analytical process trying to glean a whole lot of information out of a necessarily huge variety of recordings on an automated basis so it's neither perfect nor, in some ways, always particularly intuitive to work with at at least the lowest levels of analysis.
The up side is that it's all pretty reasonable to figure out with a little bit of time and effort, and holy shit it's such an amazing toolset. To actually be able to with a really simple script say hey this song should just have one chord and have it (with varying results) just totally work is crazy stunning stuff.
posted by cortex at 3:40 PM on September 19, 2011 [2 favorites]
I wish you could sort by date. I'm pretty sure Bolero kicked this whole thing off.
posted by Tell Me No Lies at 3:45 PM on September 19, 2011 [2 favorites]
posted by Tell Me No Lies at 3:45 PM on September 19, 2011 [2 favorites]
Also want to quickly post that this is only on a million songs -- EN has far far more than that to work with. So if you can't find your favorite thats why.
posted by brianwhitman at 3:52 PM on September 19, 2011
posted by brianwhitman at 3:52 PM on September 19, 2011
I think Pachelbel's canon predates Bolero by a smidge.
I'm mostly amazed that some of my obscure band's songs are in this database. But not any of the ones with slow builds, it seems.
posted by Casuistry at 4:02 PM on September 19, 2011
I'm mostly amazed that some of my obscure band's songs are in this database. But not any of the ones with slow builds, it seems.
posted by Casuistry at 4:02 PM on September 19, 2011
Your favorite X Y axis chart sucks.
posted by hal9k at 4:05 PM on September 19, 2011 [1 favorite]
posted by hal9k at 4:05 PM on September 19, 2011 [1 favorite]
I think Pachelbel's canon predates Bolero by a smudge.
Hey, what's two hundred years or so?
I would look for the chart on that one but the corporate firewall has suddenly decided to classify the whole site as "Weapons". The mind boggles.
posted by Tell Me No Lies at 4:11 PM on September 19, 2011
Hey, what's two hundred years or so?
I would look for the chart on that one but the corporate firewall has suddenly decided to classify the whole site as "Weapons". The mind boggles.
posted by Tell Me No Lies at 4:11 PM on September 19, 2011
With a score of 11.96, 4'33" (Cage Against The Machine Version) (Calm Before The Storm - Remix By Mr Scruff) has a rather large build.
posted by symbioid at 4:27 PM on September 19, 2011 [1 favorite]
posted by symbioid at 4:27 PM on September 19, 2011 [1 favorite]
Stairway to Heaven is # 2178.
This is like one of those okcupid data "analyses" which doesn't tell us all that much except the writer likes to play with data.
posted by bukvich at 4:37 PM on September 19, 2011
This is like one of those okcupid data "analyses" which doesn't tell us all that much except the writer likes to play with data.
posted by bukvich at 4:37 PM on September 19, 2011
If I could request a pony from Lamere, I'd ask him to please put the title and author information in plaintext
Try this.
posted by dersins at 4:37 PM on September 19, 2011
Try this.
posted by dersins at 4:37 PM on September 19, 2011
Tell Me No Lies, words are weapons, sharper than knives. Makes you wonder how the other half died... (4.24, has very small build)
posted by infinitewindow at 4:46 PM on September 19, 2011
posted by infinitewindow at 4:46 PM on September 19, 2011
most new music starts loud and stays loud, because of compression.
posted by empath at 4:51 PM on September 19, 2011
posted by empath at 4:51 PM on September 19, 2011
Also they have 15 different versions of It's Friday and no Becky Black.
And no slow build in Freebird? Oh man that is fucked up.
posted by bukvich at 5:15 PM on September 19, 2011
And no slow build in Freebird? Oh man that is fucked up.
posted by bukvich at 5:15 PM on September 19, 2011
Yeah, weird that none of those famous post-rock bands rank (Explosions in the Sky, Mogwai, etc.) An entire genre of music practically defined by trying to create the most epic slow builds ever.
Actually there's a few Mogwai songs on the list, though the first is With Portfolio which is surely cheating.
posted by eykal at 5:15 PM on September 19, 2011
Actually there's a few Mogwai songs on the list, though the first is With Portfolio which is surely cheating.
posted by eykal at 5:15 PM on September 19, 2011
Surface, by Keep of Kalessin takes the prize for linearity. That's almost a straight-line build from start to finish.
posted by rocket88 at 5:28 PM on September 19, 2011
posted by rocket88 at 5:28 PM on September 19, 2011
Also they have 15 different versions of It's Friday and no Becky Black.
the MSD was compiled in january 2011. It feels like so long ago, doesn't it.
posted by brianwhitman at 6:28 PM on September 19, 2011
the MSD was compiled in january 2011. It feels like so long ago, doesn't it.
posted by brianwhitman at 6:28 PM on September 19, 2011
I did a search for AC/DC and didn't get For Those About To Rock (We Salute You).
It has an extremely slow start and finishes with FARKING CANNONS and Brian Johnson screaming out "FIRE!" It should get a score of 99.99.
Hell's Bells also didn't show up on a search.
Your favorite slow-build band sucks.
posted by uncanny hengeman at 8:30 PM on September 19, 2011
It has an extremely slow start and finishes with FARKING CANNONS and Brian Johnson screaming out "FIRE!" It should get a score of 99.99.
Hell's Bells also didn't show up on a search.
Your favorite slow-build band sucks.
posted by uncanny hengeman at 8:30 PM on September 19, 2011
As a classical geek I'm loving that the fourth movement of Pines of Rome factors heavily in this list. Most classical musicians think of that piece as one of the loudest in the repertoire, not necessarily because it IS actually that loud (most Mahler symphonies and more contemporary orchestral music can reach the same heights, like Stravinsky, Varese etc..) but more because of the extended crescendo that is the fourth movement. I've always wished for there to be more opportunities for the "average listener" to sit inside an orchestra while it plays something immense like Resphigi. It's pretty mind-blowing.
posted by ReeMonster at 10:33 PM on September 19, 2011
posted by ReeMonster at 10:33 PM on September 19, 2011
"With a score of 1.72, A Poor Man's Memory has no slow build at all."
One of Explosions in the Sky's slow-buildiest songs (okay, granted, it's less a slow build than a long slow followed by a big jump, but that thing is made of waiting). The snare drum through the first few minutes must throw off the Stairwayerator.
posted by Mister Moofoo at 1:31 AM on September 20, 2011
One of Explosions in the Sky's slow-buildiest songs (okay, granted, it's less a slow build than a long slow followed by a big jump, but that thing is made of waiting). The snare drum through the first few minutes must throw off the Stairwayerator.
posted by Mister Moofoo at 1:31 AM on September 20, 2011
Górecki's 3rd, first movement, has no slow build at all, you say?
posted by Casuistry at 6:25 AM on September 20, 2011
posted by Casuistry at 6:25 AM on September 20, 2011
I can see a couple areas for improvement however. Right now the score is just the slope of the line fitting the curve; a separate "epicness" score would be also weighted by duration -- some of the highest scoring songs are just a minute or two long.
See: With a score of 0.49, Rez has no slow build at all
No, no and no. It's 10 minute song with *three* slow builds.
posted by Lazlo Hollyfeld at 11:52 AM on September 20, 2011
See: With a score of 0.49, Rez has no slow build at all
No, no and no. It's 10 minute song with *three* slow builds.
posted by Lazlo Hollyfeld at 11:52 AM on September 20, 2011
Hmmm...it's weird to see a couple of my albums analyzed here. Seems I'm all about the long decay (from "modest" to "substantial") rather than the slow build.
posted by malocchio at 12:23 PM on September 20, 2011
posted by malocchio at 12:23 PM on September 20, 2011
« Older The luxury vehicle of choice... for the Apocalypse | Permanent Record Newer »
This thread has been archived and is closed to new comments
posted by Halloween Jack at 2:46 PM on September 19, 2011