Sometime in the last 10 or 20 years, rock drumming has changed. Many drummers will now don headphones in the studio (and sometimes even for live performances) and synchronize their playing to an electronic metronome – the click track. This allows for easier digital editing of the recording. Since all of the measures are of equal duration, it is easy to move measures or phrases around without worry that the timing may be off. The click track has a down side – some say that songs recorded against a click track sound sterile, that the missing tempo deviations added life to a song.
I’ve always been curious about which drummers use a click track and which don’t, so I thought it might be fun to try to build a click track detector using the Echo Nest remix SDK ( remix is a Python library that allows you to analyze and manipulate music). In my first attempt, I used remix to analyze a track and then I just printed out the duration of each beat in a song and used gnuplot to plot the data. The results weren’t so good – the plot was rather noisy. It turns out there’s quite a bit of variation from beat to beat. In my second attempt I averaged the beat durations over a short window, and the resulting plot was quite good.
Now to see if we can use the plots as a click track detector. I started with a track where I knew the drummer didn’t use a click track. I’m pretty sure that Ringo never used one – so I started with the old Beatle’s track – Dizzy Miss Lizzie. Here’s the resulting plot:
This plot shows the beat duration variation (in seconds) from the average beat duration over the course of about two minutes of the song (I trimmed off the first 10 seconds, since many songs take a few seconds to get going). In this plot you can clearly see the beat duration vary over time. The 3 dips at about 90, 110 and 130 correspond to the end of a 12 bar verse, where Ringo would slightly speed up.
Now lets compare this to a computer generated drum track. I created a track in GarageBand with a looping drum and ran the same analysis. Here’s the resulting plot:
The difference is quite obvious, and stark. The computer gives a nice steady, sterile beat, compared to Ringo’s.
Now let’s try some real music that we suspect is recorded to a click track. It seems that most pop music nowadays is overproduced, so my suspicion is that an artist like Britney Spears will record against a click track. I ran the analysis on “Hit me baby one more time” (believe it or not, the song was not in my collection, so I had to go and find it on the internet, did you know that it is pretty easy to find music on the internet?). Here’s the plot:
I think it is pretty clear from the plot that “Hit me baby one more time” was recorded with a click track. And it is pretty clear that these plots make a pretty good click track detector. Flat lines correspond to tracks with little variation in beat duration. So lets explore some artists to see if they use click tracks.
First up: Weezer:
Nope, no click track for Weezer. This was a bit of a surprise for me.
How about Green Day?
Yep – clearly a click track there. How about Metallica?
No click track for Lars! Nickeback?
update: fixed nickleback plot labels (thanks tedder)
No surprise there – Nickleback uses a click track. Another numetal band (one that I rather like alot) is Breaking Benjamin:
It is clear that they use a click track too – but what is interesting here is that you can see the bridge – the hump that starts at about 130 seconds into the song.
Of course John Bonham never used a click track – but lets check for fun:
So there you have it, using the Echo Nest remix SDK, gnuplot and some human analysis of the generated plots it is pretty easy to see which tracks are recorded against a click track. To make it really clear, I’ve overlayed a few of the plots:
One final plot … the venerable stairway to heaven is noted for its gradual increase in intensity – part of that is from the volume and part comes from in increase in tempo. Jimmy Page stated that the song “speeds up like an adrenaline flow”. Let’s see if we can see this:
The steady downward slope shows shorter beat durations over the course of the song (meaning a faster song). That’s something you just can’t do with a click track. Update – as a number of commenters have pointed out, yes you can do this with a click track.
The code to generate the data for the plots is very simple:
def main(inputFile): audiofile = audio.LocalAudioFile(inputFile) beats = audiofile.analysis.beats avgList =  time = 0; output =  sum = 0 for beat in beats: time += beat.duration avg = runningAverage(avgList, beat.duration) sum += avg output.append((time, avg)) base = sum / len(output) for d in output: print d, d - base def runningAverage(list, dur): max = 16 list.append(dur) if len(list) > max: list.pop(0) return sum(list) / len(list)
I’m still a poor python programmer, so no doubt there are better Pythonic ways to do things – so let me know how to improve my Python code.
If any readers are particularly curious about whether an artist uses a click track let me know and I’ll generate the plots – or better yet, just get your own API key and run the code for yourself.
Update: If you live in the NYC area, and want to see/hear some more about remix, you might want to attend dorkbot-nyc tomorrow (Wednesday, March 4) where Brian will be talking about and demoing remix.
Update – Sten wondered (in the comments) how his band Hungry Fathers would plot given that their drummer uses a click track. Here’s an analysis of their crowd pleaser “A day without orange juice” that seems to indicate that they do indeed use a click track:
Update: More reader contributed click plots are here: More on click tracks ….
Update 2: I’ve written an application that lets you generate your own interactive click plots: The Echo Nest BPM Explorer
#1 by Stephen Green on March 2, 2009 - 9:33 am
Looks like you’re having fun over there!
Did you try the python module for Netbeans, or are you still vimming it?
#2 by plamere on March 2, 2009 - 9:34 am
Still viming, but I may get a chance to try the netbeans/python module today (and thanks for that tip).
#3 by Brian Utterback on March 2, 2009 - 9:54 am
But you haven’t really proved your supposition. I grant that it is highly unlikely that a drummer without a click track can produce such flat graphs, but who knows? When were click tracks invented? You should try to find the flattest plot for a piece of music prior to that time. Also, are there any drummers that use a click track but use their own judgment and deviate from it?
It is interesting that your plots show that the drummers can hit the beat to within less than a millisecond. It is commonly known to horologists that humans can match a beat highly accurately even though we are pretty poor at judging durations. This shows that phase-locked loops pre-date electricity.
#4 by Danno on October 18, 2009 - 5:36 am
Hey Brian U, Boss has been making metronomes with earbuds and LED’s since the late 70’s – early 80’s. Mechanical metronomes have been around since the early 19th century, and musical compositions since that time have included markings that correspond to a metronomes BPM.
Your second paragraph brings up a good point: grab a Robert Johnson record and “graph” him. His time signatures change sometimes every measure!
Like this article. Your next subject should be Dragonforce!
#5 by plamere on March 2, 2009 - 10:02 am
@brian – indeed, this is not good science – to do this properly, I should verify my hypothesis with a large number of positive and negative examples. Indeed, there may be some hyper accurate drummers that look like click tracks and conversely there may be some click tracks that intentionally vary from the perfect tempo. So… consider this an exploratory venture into click track detection. Certainly the qualitative difference between plots points to something interesting.
#6 by Sten Anderson on March 2, 2009 - 10:12 am
Our drummer for Hungry Fathers uses a click track both recorded and live. I’d be curious to see if we came up correctly as part of the control group.
#7 by brian on March 2, 2009 - 10:22 am
let’s hear ringo’s sloppy drumming applied to “hit me baby one more time”!
#8 by S on March 2, 2009 - 2:09 pm
Couldn’t “sloppy” drumming also be cleaned up after the fact? Probably easier to use a click track and make it steady from the beginning if you so desire, but I’m sure it must be possible to get similar effects in post production.
#9 by Chris Cothrun on March 2, 2009 - 2:58 pm
The linearity of the clicktrack plots is surprising to me. Cakewalk’s products, the audio programs I’m most familiar with, easily allow the producer to change the tempo of the MIDI tracks that would form a basis for click tracks. I’m also aware of the influence that slight tempo changes can have on the emotion conveyed by the music and I’ve always assumed that these features were being used by the ‘pros’. After all, the effect has a name: rubato
I’d love to see more analysis of drummers. The Kashmir graph fits well with what many describe as his signature ‘behind the beat’ style of playing.
#10 by tedder on March 2, 2009 - 3:37 pm
Brilliant. I’m hoping you analyze more music- it’s interesting to see how click tracks and auto-tune have shaped modern (commercial) music.
#11 by orbnotedgo on March 2, 2009 - 3:38 pm
Excellent analysis! My one bone to pick:
“The steady downward slope shows shorter beat durations over the course of the song (meaning a faster song). That’s something you just can’t do with a click track.”
Hmm – well, that would assume that a click track stays at a fixed tempo. While it appears that many of your examples use a steady tempo, a click track can actually be assigned ahead of time to change tempo at various points while the recording is underway. I’ve often played to a click track that varies tempo throughout a song, specifically at verse/chorus transitions. As well, most modern DAWs can apply curved automation to tempo so that speeding up and slowing down seems considerably more natural. Playing with a click track can be used for effect, too. When done well, the result can positively change the attitude of the recording.
#12 by plamere on March 2, 2009 - 3:43 pm
@orbnotedgo – thanks – I had no idea that click tracks were so sophisticated. I imagine it can make it much more difficult for the drummer to follow along in a live situation – and if you get out of sync it must be disastrous!
#13 by tedder on March 2, 2009 - 3:55 pm
Hmm, some followup.
nitpick: Your graph for Nickelback still says Metallica: Enter Sandman at the top of the chart, which is slightly confusing.
More substantial suggestions: It would be interesting to take some of the same artists and compare their work over time. For instance, the Metallica track is an old track, and the Green Day track is very new. Find some of their early, mid and late work and see if it all matches, or if artists appear to be using click tracks more now. Second, pull live tracks from the bands (official and unofficial), it’d be interesting to see if bands use click tracks more or less when they are live.
#14 by plamere on March 2, 2009 - 4:06 pm
@tedder – thanks for the nitpick – I’ve fixed the chart labels. Excellent suggestions about doing a longitudinal study of a band to see how things have changed over time. Likewise, comparing live and studio recordings is a great idea. If only there was more time in a day!
#15 by tedder on March 2, 2009 - 5:05 pm
Is your code something you could share? If so, perhaps there are individuals (such as myself) that be interested in running it to help find trends.
#16 by plamere on March 2, 2009 - 5:11 pm
@tedder – Sharing code? you bet. The bulk of the code is already in the original post, the rest of the code is just setting things up. The actual analysis is all done using the echo nest web service that anyone can use. You just need to get a key (for free) at developer.echonest.com. One of the things on my task list is to set up some pages that show examples like this on the echo nest site, where people can come and download the code and run it. Look for that soon (like in the next week or so). If you want to get started earlier than that just email me at Paul@echonest.com and I’ll email you the 20 lines of python. I think it would be neat to see data for all sorts of bands at different stages in their careers.
#17 by TClark on March 2, 2009 - 6:38 pm
Before people jump over into “See? Who needs click tracks anyway?!?!” territory, I think it worthwhile to consider the Y axis of these plots. Consider Lars Ulrich and Ringo — their variations are around 0.01 second.
They’re keeping DAMN good time when the difference is measured to be around 10 milliseconds.
#18 by Hal on March 2, 2009 - 8:49 pm
interesting stuff. see link below, basically are modern drummers unconscious, auto-corrected by software or just very good?
“The average reaction time is between 200 and 270 milliseconds, although it can be even faster among athletes.(2) Interestingly, it has been conventionally accepted that it takes 500 milliseconds for conscious processing to occur, which manifests as cortical or thalamic activity.(3) Recently, this notion has been challenged by S. Kitazawa whose research has shown conscious sensation takes only 80 milliseconds after the stimulus, and at 500 milliseconds this sensation is “localized in space.”(4) This difference has some very provocative implications toward consciousness and what behaviors are constituted as “conscious.””
#19 by Jonathan Foote on March 3, 2009 - 12:31 am
As an aside, there is a musical urban legend about a session drummer (Steve Gadd?) who was so rhythmically precise that engineers could disconnect his headphone feed while overdubbing (so he couldn’t hear the tempo of the tracks/clicktrack he was playing along with) and he still kept perfect, solid time.
Another interesting thing to look at might be “Strawberry Fields Forever” which was famously spliced together from two different takes.
#20 by Julius Davies on March 3, 2009 - 2:12 am
Would it be too crazy to suppose that “Hit Me Baby One More Time” doesn’t use a click track, nor a drummer, and that the drum part is sequenced?
(Same for George Michael I suspect! I swear he is the master of the 80’s drum machine.)
#21 by Julius Davies on March 3, 2009 - 2:21 am
e.g. Freedom 90 by George Michael
#22 by ken on March 3, 2009 - 2:46 am
as someone who played [gtrs] on countless albums, ads movies and similar through the 80’s [when click tracks become rife in recorded music], you’ll find it’s been the overwhelming convention from then till now to use a click or some synth pattern to track songs to. i produce a lot of music now, and my opinion is it’s neither good nor bad. in fact I’d bet it’s saved many a band who were crap at playing together but got forgiven it when playing live cos the songs had enough going for them. my experience of most humans is they’ll forgive massive amounts of slipping around in time and between players when it’s live and the vibe is high, but can easily sense things as ‘sloppy’ or ‘its falling apart’ when they hear a recorded performance in their iPod ie not at a gig with the attendant rush.
tho popular trend is towards uncontrived ‘performed now’ immediacy on recordings most people wouldn’t cope with how most artists sound without a lot of help …. i’d bet you the most common aid is still a click track. … click tracks save a shitload of pain for most artists esp when overdubbing all the niceties later and give countless other freedoms like you can replace certain drums [snares and kick replacement is rife in lotsa rock & metal for ex] or add all kinds of electronic sonics later with the computer thats locked to yr click track [and sounds ‘right’ since the band doesn’t stray from predictable time too much.
you’ll also find a lot of the better session guys from the 80’s/90s particularly the best of the US ppl could sit like a metronome without one.
#23 by Ben on March 3, 2009 - 2:48 am
I can’t imagine click tracks that change tempo would be any harder for a drummer to follow than click tracks that don’t. As you’ve shown they’re already not doing what comes naturally.
#24 by Bedroom on March 3, 2009 - 3:23 am
I seem to recall a home video of Metallica (“A Year And A Half In The Life Of…) where we see Lars discussing the tempo changes in Enter Sandman, saying something like “we’ll do the intro at 125BPM, the verse at 121BPM…”
I think Lars really did use a click track, but with a moving tempo. I did the same thing for a song I recorded with my old band a long time ago. I used a MIDI sequencer to create a click track for my drummer with a couple measures at 170BPM, followed by a verse at 145BPM, chorus at 155BPM etc.
#25 by robertwb on March 3, 2009 - 3:26 am
Humans are capable of processing such tiny amounts of time, for example the difference between aspirated (p,k,t) and voiced (b,g,d) consonants is a matter of 10s of milliseconds, and humans are able to perceive such differences shortly after birth. Using a click track doesn’t measure reaction time; a good drummer can be expected to keep a pretty consistant beat. The click track prevents her from drifting as the song progresses.
#26 by Tom on March 3, 2009 - 3:27 am
Something is not right with your code – the mean of the deviation from the mean should be exactly zero, but for many of these graphs it is obviously not. The most obvious ones are for Nickleback and Hungry Fathers – the ‘deviation from the mean’ is less than zero for the whole duration, so you must be mis-calculating the mean somehow.
#27 by Loren on March 3, 2009 - 3:32 am
Reaction time is usually measured in response to unexpected stimuli. Keeping a beat is the exact opposite effect. The brain is incredibly good at forming models and synchronizing motor functions to expectations. Think about everything involved in a simple act like walking. It’s no surprise to me at all that even a mediocre drummer can keep a beat to within a few milliseconds for short periods. What seems to challenge most drummers is keeping the tempo during fills, and maintaining the tempo over longer periods. Very few drummers have innately perfect time, but 30-50 milliseconds at 144bpm is about a 32nd note duration, which is enough to sound ‘off’ to the average listener, and is very obvious to experienced listeners.
#28 by Soteriologist on March 3, 2009 - 3:37 am
Ok… You need to test some Neil Peart (from Rush) try the song “Tom Sawyer” from the album “Moving Pictures”, Gavin Harrison (from Porcupine Tree) try the song “What Happens Now” from the album “Nil Recurring”, Mike Portnoy (from Dream Theater) try the song “The Dark Eternal Night” from the album “Systematic Chaos”.
The only problem with picking a really good drummer is… what if they’re so good that it LOOKS like they’re using a metronome, but they’re really not. Neil Peart is known as the “human metronome” for goodness sake. We’ll see. Or maybe an old Dream Theater track would be better for Portnoy because he might not of used one back then. You could try the song “Ytse Jam” off of Dream Theater’s first album.
It’d also be cool to see some other goodies up there, like Terry Bozzio, Mike Mangini, Danny Carey, Vinnie Paul, Jimmy Chamberlin, Phil Collins, Carter Beauford, Dave Grohl, Keith Moon and we can’t forget good old Buddy Rich!
What’s really be cool is to get an averaging/analysis of an entire body of work from each artist (or as close to an entire body as possible). Like all recorded works by Portnoy, etc.
#29 by Michel on March 3, 2009 - 3:57 am
There’s probably an even more accurate way of detecting click tracks: instead of plotting the beat durations individually (or averaged over short periods), you could use the beat times measured from the start of the song, and compare this to a click track with the same average beat duration. As the song goes along, the click track drummer will stay in sync, even if he’s a really bad drummer, while even the best drummers will slowly diverge as all the small deviations add up.
The bad drummer with the click track may miss individual beats by many milliseconds, but this will always be corrected later on as the click track drags him along. You were already seeing this when you started averaging short durations instead of using individual beats, and the effect will be even more pronounced when you measure from the start.
#30 by Davey on March 3, 2009 - 4:03 am
The speed-up/slow-down tempo thing you can see on the Breaking Benjamin track can be done with a click track, using most music software used for live work – e.g. Propellerhead’s Reason or Ableton Live. You can alter click tracks by fractions of BPM, and stuff like loops and samples are altered to fit (without changing pitch). Neat!
In fact, if you want to finish a track that changes tempo on click, it’s best to control the click tempo and keep it going – it can be rather unweildy to pick up a click at a different tempo when you’re playing.
#31 by Am on March 3, 2009 - 4:19 am
For the vast majority of pro drummers a click track isn’t really chosen in order to keep a drummer in time – they’d be pretty rubbish if that was required and the human ear as the article suggests quite likes a bit of ebb and flow in tempo. But it isn’t for sound-editing as the article suggests because if you need to nudge a beat that’s a bit off or similar you have very granular resolutions down to 100ths of a beat to which you can ‘snap’ to or indeed you can just do it manually. Playing to a click doesn’t add to that at all.
By far and away the most often seen scenario is that the use of click tracks is dictated by tracks which use sequencers for keyboard parts, bass lines, percussion etc need a drummer to use a click track to stay in time with the pre-programmed parts. And of course it is dance music and Britney Spears type acts that require that… whereas yeah Lars can take a reference tempo at the beginning of a track and then let the music breath as he’s feeling it.
#32 by Michael Reed on March 3, 2009 - 4:50 am
What about the effects of cut and pasting and looping? I presume that a single play through of a bar that is looped would give a flat group?
In addition, when I’m recording, I sometimes record a single verse and chorus of some parts. So, for example, the drums, guitar, synth etc for verse 3 are identical copies of verse 1.
Another thing that occurs to me is that with some music, particularly pop, the drums might not be recorded first. So, the drummer might be playing along to a sequenced synth line rather than a tick per se.
#33 by mclaren on March 3, 2009 - 4:58 am
Kudos for doing the hard work to provide hard quantitative data here. All too many musicians start out with an hypothesis, say to themselves, “Boy, that sounds logical,” and then simply accept it without any sort of objective test.
Several points. First, to nail this down you need to do a double-blind test. Get some musicians to lay down tracks both with and without a click track and give you the recordings. After you analyze ’em without knowing which was which, they then reveal which recordings used a click track.
You claim that accelerando or decelerando can’t be done with a click track. Alas, not so. Cakewalk (and many other MIDI sequencing packages) have a tempo change feature. You can step-enter beats and then speed them up or slow them down over the course of a MIDI sequence by any desired percentage. A drummer playing to such a click track would produce an accelerating or decelerating beat.
Lastly, current DAW software like Nuendo and Pro Tools allow audio recordings to be treated like MIDI sequences. I.e., recording drum hits in the audio track can be auto-corrected to within a specified percentage accuracy of the beat. This is particularly easy to do with drum tracks, since the drum strikes are typically percussive timbes with short decay times, so the entire audio event can easily be moved forward or backward in time by DSP software. Are you sure that isn’t what’s happening here?
#34 by elb on March 3, 2009 - 4:58 am
Have you checked out Sonic Visualiser with the Queen Mary Tempo Tracker Vamp plugin?
#35 by Mike on March 3, 2009 - 5:00 am
Music wouldn’t be about reaction time though. A drummer isn’t reacting to a beat, he is predicting it. If he was thinking about it he’d be too slow, as you said.
#36 by plamere on March 3, 2009 - 5:11 am
@tom – “Something is not right with your code -” you are absolutely right, there was a bug when I generated the plots – I was dutifully calculating the average beat duration but instead of subtracting all durations by this average, I was subtracting the duration of the last beat. I fixed the bug before I posted the code, but since it didn’t affect the shape of the plots (and I’m rather lazy), I didn’t bother to fix them. Anyway, you have a good eye for data!
#37 by Dude on March 3, 2009 - 5:15 am
The White Stripes, she’s pretty renown for her off-beat drums … would be nice to see what it looks like.
Excellent initiative, btw. I wonder how the music industry is going to counter this. Trust me, they will.
#38 by Carl Blomqvist on March 3, 2009 - 5:23 am
Have you tried analyzing bands that play really fast and heavy music, with lots of tempo and time signature changes?
Try these bands if you can:
The Faceless – Fast, many changes in time signature
Periphery – Many uncommmon time signatures.
I would also like to add how obvious it is that nickelback uses a click track. There is a song out there wich is simply two of their songs hard panned Left and Right. It’s called How You Remind Me of Someday, and the only place I found it (didn’t look to hard) was here:
Just ignore the video. Both songs are pretty much built up the same way, aswell as having the exact same BPM and key. In the youtube video there is no panning though.
#39 by Gio on March 3, 2009 - 5:26 am
Could you also run “Beat It” through your algorithm? There’s a famous story here about how drummer Jeff Porcaro recorded the whole thing without a click track and still made it sound like a drum computer… Thanks!
#40 by Jean-Francois Im on March 3, 2009 - 5:29 am
Michel: I’m not sure that I fully understand what you’re saying. Are you suggesting to take the average beat duration from the beginning of the song(say the first 16 measures) and extrapolating the rest of the song from there?
Also, congrats on being slashdotted Paul. :)
#41 by plamere on March 3, 2009 - 5:33 am
Jean-Francois – I think what Michael is saying is that since the click track won’t vary, if you look at the absolute time of the final click modulo the tempo will be the same as the absolute time of the first click modulo the tempo. It is a good idea.
#42 by t o b e on March 3, 2009 - 6:27 am
Cedric Sharply would be a good one to check. I don’t think he used a click but his almost machine like. He’s the drummer on Gary Numan’s Pleasure Principle.
#43 by BC on March 3, 2009 - 7:53 am
Try it with something off Metallica’s Black Album — legend has it that Lars’ desire for perfection led to the album’s being delayed. This was the first time I had heard of a “click track” being used, FWIW.
#44 by Gary on March 3, 2009 - 7:56 am
I was wondering which Metallica song you tested? I doubt the Lars of old used one, but in the making of their newest album, he is definitely trying to hear something to time his drumming.
Could you test out something from the Death Magnetic album and see if he uses one in that one?
#45 by plamere on March 3, 2009 - 8:18 am
@gary – the metallica song was Enter the sandman. BTW, the song titles for all plots are shown in the title of the plot itself.
#46 by chris on March 3, 2009 - 8:34 am
Try doing stairway again with the code set up to increase in tempo based on the begining average tempo and end of the tracks average tempo… I’d like to see how close they got to a perfect through the whole song.
#47 by Robert on March 3, 2009 - 9:07 am
Click tracks become inaudible to the drummer with a comfort level to self adjust the “pulse”. Run the program on material by Dave Weckl and Steve Smith and you’ll find that they are capable to playing to clicks and yet perform subtle adjustments to the notes parking themselves before, on and after the beat or click within a given measure.
As mentioned above more sophisticated recording and performance software prorams can add “swing” and decelleration/accelleration to a rhythm click track.
I couldn’t agree more with you that some music suffers the sterille effect with a locked in drum track; not dance music, but rock does.
#48 by john on March 3, 2009 - 9:18 am
Keith Moon was obviously playing with synths for Baba O’Reily, Won’t Get Fooled Again, and Who Are You. It’d be interesting to see the chart for these. I would be especially interested in the live “The Kids Are Alright” versions of the first two songs, as he was having a lot of trouble keeping tempo at the concerts.
#49 by Michael Chaney on March 3, 2009 - 9:32 am
I second what Julies Davies said – some of what you’re seeing is likely just sequenced, which would be a step above the click track.
#50 by picoshark on March 3, 2009 - 9:39 am
MIDI sequencers and DAW have been able to ramp tempo since 1986 (that I have experienced), perhaps earler.
#51 by ken on March 3, 2009 - 9:41 am
the best drummers – and other musicians – can float round a click track, rythmn loop, any loop, any drum machine… in fact a lot of great music had some time “aid” like the previous and the players brought the performance to life by minutely breathing from centre time to ahead or back of the beat.
this is so irrelevant. sometimes absolutely robotic metronomic time will be what the music needs … other times it will stifle the song/mood/whatever.
it’s just another aid to trying to make good music. just like an amp aids the guitarist by opening up other sonic territory… or a capo helps a guitarist play things across many keys that would be otherwise difficult or impossible….or a synth program that plays a whole complex rhythm on one key. or a sampler turns a vocal into a tripped out hooky soundbite [ Missy Elliots ‘Work it”].
if you didn’t know about it – just get used to it. it doesn’t kill music any more that it guarantees cool music. it just is. the humans using it are responsible, for the better of the music, or not. just as most people just like it, love it or don’t, regardless of whether the bass player had stickers on the frets to remind herself of where to play or not.
#52 by Murple on March 3, 2009 - 10:22 am
Try running some old Art Blakey through this. Arguably the greastest drummer of all time. Lots of his stuff may have too many tempo changes, but I’d be curious how close to a straight line he gets on tracks where he plays a fairly steady tempo.
#53 by Stat on March 3, 2009 - 10:30 am
How about some old school hip hop from Boogie Down Productions. Part the song called Poisonous Products has the verse “It’s kind off beat… yea I know it.”
#54 by Mike on March 3, 2009 - 10:37 am
Very cool! I wonder, could you do something analogous to
#55 by Francis Van Aeken on March 3, 2009 - 11:15 am
You can use my tool ‘JacksonDJ’ to easily find out whether a song has a fixed tempo. Disclaimer: there is currently no support.
#56 by brian on March 3, 2009 - 11:16 am
mike: very likely. the remix/analyze data outputs pitch (chroma) ratios as well, you could watch for variance among the different bins to see if a pitch corrector is in place.
#57 by Jeff on March 3, 2009 - 11:53 am
Bill Bruford has said that in his Yes days, there was no attention paid at all to steadiness of tempo; it just wasn’t an issue. But of Ringo, Paul McCartney has said that Ringo’s ability to hit and resume a given tempo reliably made it possible for the Beatles to use radical edits without ruining a mix.
While interesting as an exercise, I feel like this process isn’t really useful for evaluating drummers qualitatively unless the drummers are playing by themselves. Personally, I feel like in group playing – which is what these plots are more representative of – the drummer is part of a consensus. I would contend that the drummer does not necessarily have a unique role in that consensus; it’s just that his or her expression of it is more uniformly obvious.
#58 by plamere on March 3, 2009 - 12:00 pm
@jeff – excellent point about the beat as a consensus among the players. The Weezer plot is particularly telling. The big peak on the right occurs when when the drums stop, and the guitar a bass play alone for 20 seconds. The tempo sense is out of wack. But when the drums return, things get back to where they should be, more or less.
#59 by burninator on March 3, 2009 - 12:00 pm
Dave Grohl’s comment on using click tracks (from a 2001 interview):
“We do use click tracks sometimes. I remember the first time someone advised me to use the click track I was absolutely heartbroken. I’m sure any time a producer says, ‘You might want to think about using a click track on that’, your ten or 15 years worth of experience go right out the window and you feel horrible. So we were recording Nevermind, laying down the basic track for ‘Lithium’, and we had maybe gone through three or four takes and Butch Vig brought us in and said to me, ‘You might want to think about using a click track’, and I immediately felt worthless and horrible. But I learned that as long as you can play within the click track you can still achieve a natural feel. I’m convinced and I used to be a purist and think, ‘Jon Bonham never used a click track!’ But I think it depends on what you are trying to achieve. If there’s something that you want to sound completely chaotic and has not only a dynamic but a tempo dynamic that moves up and down – okay. If there’s something that you want to be completely solid – okay. I’m not opposed to trying new things. I’m not into drum loops or drum machines for this band, but I think click tracks can be helpful.”
#60 by foo on March 3, 2009 - 12:11 pm
It’s an arms race. Now all the smart music producers are going to go back and analyze all the best old songs using your techniques, find out the common tempo pattern variations, and adjust the click tracks to provide a more “natural” feel for the structure of the song.
#61 by TRK on March 3, 2009 - 12:16 pm
I have sat in studios and watched engineers spend a lot of time in post production editing drum tracks to get the tempo and feel they want. So I’m not sure you can say this method will always show if the drummer originally played to a click track.
#62 by Who moon Fan on March 3, 2009 - 12:31 pm
Yeah, please someone do a Kieth Moon for Mobile, and maybe bargain, which was a lot smoother.
Kieth – not the best but a very interesting drummer!
#63 by Landros on March 3, 2009 - 12:40 pm
I have worked in a well working studio, and indeed, some click tracks are made to have accelerations and slow downs on good productions to keep “life” in the beats.
I would like to say that also many bands now use drum programs such as BFD or Addictive Drums instead of recording real drums, this is a great gain of time and has much larger mixing possibilities. I am quite sure the drums on that Britney Spears song are not real drums, but simply sequenced samples.
Any ways it’s very interesting to see the variations of those drummers, and Ringo’s precision in his accelerations and slow downs.
#64 by David in Boston on March 3, 2009 - 12:48 pm
Just for fun, how about showing a Rick Astley tune… After all, we have all been Rick Roll’d (http://en.wikipedia.org/wiki/Rickrolling) at one point or another…
#65 by SpamAnd Eggs on March 3, 2009 - 12:53 pm
Interesting stuff… but one thing I’d like to point out (yes, I am a drummer…) that one thing I believe is much much more important than tempo is the feel or ‘groove’ of a song… some songs just ‘lock in’, and some just don’t, but if it feels good, it’s perfectly OK… one thing that drives me absolutely nuts is other musicians who get so completely anal about tempo, that they forget to just play the music… Click tracks? – producers like them (cheaper and easier edits with digital tools), but most drummers I know don’t like them, since it’s more difficult to play with feeling when most of your concentration is on the click…
#66 by Patrick Richardson on March 3, 2009 - 1:27 pm
Hey Paul. Nice seeing you at ISMIR 2008.
Nice article. I was surprised at the degree of deviation that was evidenced by your analysis in some familiar songs. Kinda speaks to how forgiving the psychological elements of our music can be.
I’ve been working with and without click tracks in my studio for years now. I love it and I hate it, because it allow us to compose our songs after we start tracking. Even though music production software synths and loops allow us to compose-through-producing and re-imagine (remix) our ideas in completely non-linear ways, “to click or not to click” remains a singular yoking initial decision. As a drummer, I’ve felt the need to become more aware and involved in the production process.
#67 by Jojhn La Grou on March 3, 2009 - 2:26 pm
Paul, great stuff.
On a historical note, the first click track (actually servo track, but same concept) was recently found in France – it predates Edison’s first sound recording by many years. The Lawrence Livermore guys played it back, here:
The servo track sat underneath the audio track. There was an image of this on-line, but it seems to have been removed due to a copyright issue.
#68 by angryechidna on March 3, 2009 - 2:31 pm
Really interesting stuff.
Little more food for thought: Greenday (not sure which song you ran this on) often sticks to more straight ahead punk beats whereas other bands (Led Zep for instance) often played more of a blues-rock style. I’m no advanced drummer, but I wonder how much easier it is to keep a tight beat between different styles of music?
#69 by Jason Forester on March 3, 2009 - 3:52 pm
Very nice! Just a few thoughts:
1) it would be very interesting to see the effects of quantization when plotted – does it take a track with human variance and make it appear to be done to a click track?
2) the opposite is also interesting – many drum programs can now add so-called ‘human feel’ by varying the beat slightly in emulation of humans who keep a ‘mean’ time vs. an ‘absolute’ time through the song.
Of course whenever people see something new they have a hundred experiments for the inventor to run – nice work regardless. Keep up the good work, look forward to more,
#70 by Antti Rasinen on March 3, 2009 - 3:57 pm
Some notes about the code snippet:
1. You can do tuple unpacking inside for loops. Instead of “for d in output” you can write for time, avg in output: …
2. Multiple assignments are useful when you have short variable names and similar values to insert: time, sum, output = 0, 0, 
3. sum is a sort of a bad name, since there is a builtin by the same name. But it turns out you don’t really need sum, because base is pretty much audiofile length (in seconds) divided by the number of beats (or len(beats)).
4. runningAverage and avglist might be separated into a FIFO buffer class. If you have Python 2.6, then this class is pretty much collections.deque with maxlen=16. Deque is also slightly better than list when you’re doing a FIFO buffer.
from collections import deque
def __init__(self, size):
self.buf, self.size = deque(), 16
def append(self, x):
if len(self.buf) > self.size:
def mean(self): return sum(self.buf)/len(self.buf)
and later in the code you’d write buf = FIFOBuffer(16); …; buf.append(…); avg = buf.mean()
#71 by plamere on March 3, 2009 - 4:02 pm
@antti – i love the internet. Thanks for the great code review.!
#72 by d mann on March 3, 2009 - 3:59 pm
“Another interesting thing to look at might be “Strawberry Fields Forever” which was famously spliced together from two different takes.”
Though the two takes were initially at different tempos, in different keys, and one was sped up, the other slowed down to make the splice work…it would be interesting but it wouldn’t show whether Ringo could “lock in” a tempo.
#73 by Vudu 12 on March 3, 2009 - 4:46 pm
Very cool! But what is “0.01”? Percent, seconds/bar, ?
It’s really hard to know what we are talking about without this info. I can’t believe Stairway only has 0.5 s/bar variation between the beginning and end.
The first 10 seconds of the songs would be interesting to see as well.
Songs with tempo movement are not very well accepted in todays digital world, it sounds “old” and maybe even wrong or sloppy. Like “get a drummer who can play”.
Digitally changing tempo, ramping or jumps for the chorus, end of song, are digital approximations that may help add emotional variations but are still pretty weak compared to live drumming. But subtly done may be better than nothing.
#74 by Pete Best on March 3, 2009 - 4:56 pm
>The difference is quite obvious, and stark.
In this case it’s also quite Starkey. :)
#75 by Adam Williamson on March 3, 2009 - 5:08 pm
Just to agree with Landros – I would expect Britney’s percussion tracks are mostly computer generated using a sequencer, I doubt there was a real person hitting a drum kit (electronic or acoustic) at any point. It doesn’t fit the sound of most of her songs.
#76 by oldschool Houserock on March 3, 2009 - 5:10 pm
besides a drummer playing to a click, editing software where a tech corrects each drum hit is probably more the heart of the matter. if a kick drum is behind, they simply nudge the wave in time to make it more drum machine like. If you watch metallica’s SOME KIND OF MONSTER, there’s a scene where they’re editing drum takes to be in time with the click.
Most drummers cant play perfectly to a click anyway. They get close and someone edits it to be a perfect take.
#77 by Vudu 12 on March 3, 2009 - 5:15 pm
I just miraculously remembered a reggae track that sounds very solid and stable, but I had accidentally looped it, and was shocked to find that the beginning is very slow compared to what it ends like.
It is Gregory Isaacs, Nigh Nurse album (1982) and ironically the song is called “Cool Down The Pace”. I’d really like to see it graphed please!
#78 by Mark on March 3, 2009 - 5:38 pm
Something else that this doesn’t really account for, though, is the current use/abuse of tools like ‘beat detective’. Essentially, quantizing of recorded audio tracks (much like one would quantize MIDI information).
Once a drummer’s tracks are quantized, they usually finish it off by sample-replacing all the sounds.. making the drummer a human stand-in for a drum machine.
#79 by Andrew on March 3, 2009 - 5:39 pm
Very interesting article… You should try a song with a lot of time signature changes like “Tool – Schism”
#80 by dhobsd on March 3, 2009 - 6:00 pm
I decided to take a look at Between the Buried and Me’s drummer on Prequel to the Sequel (off of their album Colors). This song has quite a few tempo changes and time signature changes; I was curious as to how this would look, since I know for a fact that he does use a metronome while drumming.
This proves the point of “Yes you can do this with a click track”
I made the live graph because I also happen to know that he wears a metronome while playing live. The graph here does seem to confirm that he was wearing one for this concert (and Christ, his accuracy is good).
However, it does go to show that the code doesn’t take tempo changes into account very well. I haven’t looked at the API much yet, but it’d be nice to be able to do some smoothing based on this.
#81 by plamere on March 3, 2009 - 6:24 pm
@dhosbsd – nice job with the plots. In fact, the API will help you sort out the tempo changes. There are api calls that help you sort out the song structure – sections, bars, beats and tatums are all available. It is likely that each tempo change would be associated with a different section. So it would be very easy to modify the code to calculate the baseline average for each section instead of for the whole song. That is, of course, an exercise left for the commenter ;)
#82 by Tim McNerney on March 3, 2009 - 6:20 pm
I see that most of your analysis has been of rock, metal, and pop. Back in the early 1990’s, I noticed that country music drummers had this amazing steadiness. Ever since I have wondered if click tracks were behind this. There are a number of tracks on Nanci Griffith’s “Late Night Grand Hotel” that come to mind. The opening track, “It’s Just Another Morning here” is probably as good as any for you to test. If you are interested in applying your software to this track, I’ll send you the 99 cents to buy it from iTunes. Alas, I don’t have the CD anymore.
#83 by dhobsd on March 3, 2009 - 6:30 pm
Sure, I’ll probably actually do that. Right now I’m having an issue with the API not working (getting list index out of range on echonest/web/util.py:53), but feel free to mail me outside of the comments (I’m assuming you get to see my email address when you go through the comments?). This is actually a subject I have a lot of interest in (even if it’s just academically; I don’t have any practical use for it). I’ll post some updated code later tonight / tomorrow.
#84 by synthdave on March 3, 2009 - 6:54 pm
Two other technologies that changed music in the last decade are beat correction software and sampling.
It is not uncommon for a drum track to be sliced up, conformed to a tempo grid, and then have all the individual drum hits replaced with samples.
While sampling is relatively new, beat correction, goes back way before music software. An engineer would sit next to a reel of 2″ tape with a ruler, a grease pencil, a razor blade and splicing tape for days and would essentially do the same thing that beat correction software now does in seconds.
#85 by Tim McNerney on March 3, 2009 - 7:16 pm
Actually, I gather there is a fair bit of interest in tempo analysis. NPR interviewed a researcher who found a correlation between popularity of songs with low tempo variation and times of high economic anxiety, and vice versa. I think he did this over the past 50 years, which of course included several decades before drum machines.
#86 by Matt on March 3, 2009 - 8:58 pm
It does not look like this would be easy to try and get something running on my own to try and break this. I wanted to try some really fast tempo musing that uses a human drummer (so a lot of dance/techno is out). In that vein, maybe try something like Dragonforce to see what happens.
#87 by Keith Kehrer on March 3, 2009 - 9:33 pm
How very interesting. I have played in boatload of bands and have also recorded a lot of electronic dance tracks and I have to say that it is really personal preference and also the style of music that would dictate whether you used a click track or not. Electronica thrives on that steady click, but I love the ebb of flow of playing with an expressive drummer. I am sure U2 uses a click and I think Lars should use one. :) I would be interested in seeing Motown drummers analyzed because they always seemed to be steady tempo-wise.
Thanks for doing all that. Food for thought.
#88 by ken on March 3, 2009 - 9:43 pm
>>” If there’s something that you want to sound completely chaotic and has not only a dynamic but a tempo dynamic that moves up and down – okay. If there’s something that you want to be completely solid – okay. I’m not opposed to trying new things. I’m not into drum loops or drum machines for this band, but I think click tracks can be helpful. [Dave Grohl]”
perfectly put Dave – thank you burninator
quote: [I’m no advanced drummer, but I wonder how much easier it is to keep a tight beat between different styles of music? angryechidna ]
i guess yr an aussie??… in my experience how much easier it is for any drummer [or any other instrument player] depends on the players innate sense of strict time; their skill levels on their instrument [chops]; the complexity of the patterns theyre attempting on their kit; the tempo [really slow tempos are the hardest to keep dead solid IME] and perhaps most importantly – how strong with time the rest of the players theyre playing with are. The answer for most players is “NEVER EASY” which is why click tracks at least of all the time aids ARE so widely used.
if yr a player try it. get a metronome or a click tk: play something dead simple and record yrself w the click for a minute or so. listen back. then try it with two friends. it’s harder with others involved, and i’d bet for most players even solo you’ll drift within 3 or 4 bars even on the simplest patterns. then try it at 70bpm.
it took me 2 or so years of focussed practice and about the same again in busy studio time to get to a point where i was pretty much able to sit front, back of, or centre time on a click with authority in just about any tempo. this was after i’d been playing [gtr ie] for about 12 yrs or so.
add a bands worth of ppl playing around you in the headphones and depending on the quality of the monitor mix and that system – and it’s near impossible to stay full locked unless everyone else is VERY sharp and good at it.
again: clicks, loops whatever are just another tool. blaming the click tracks [for sterility or other negative] is just as idiotic as blaming the choice of recording tape or hard drive or drum kit the drummer had…. they’re all just choices of tools for getting a result.
#89 by brian on March 3, 2009 - 9:45 pm
tim: that was another echo nest analysis customer, Philip Maymin
#90 by ROBERT on March 3, 2009 - 10:01 pm
Ouch…….My brain hurts……..thanks a lot!
#91 by Void Main on March 3, 2009 - 11:58 pm
You left a few lines of code out of your code listing but I did get it to work. Would you mind posting your gnuplot source? Here’s my “click.py” listing:
——– cut ——-
import echonest.audio as audio
usage = “””
python click.py EverythingIsOnTheOne.mp3
audiofile = audio.LocalAudioFile(inputFile)
beats = audiofile.analysis.beats
avgList = 
time = 0;
output = 
sum = 0
for beat in beats:
time += beat.duration
avg = runningAverage(avgList, beat.duration)
sum += avg
base = sum / len(output)
for d in output:
print d, d – base
def runningAverage(list, dur):
max = 16
if len(list) > max:
return sum(list) / len(list)
if __name__ == ‘__main__’:
inputFilename = sys.argv[-1]
——- cut ——–
I’m sure the formatting will get lost so:
#92 by Scott on March 4, 2009 - 12:10 am
The soul of the machine is a restless, natural engine!
Just say NO to the CT!
#93 by Richard Bronosky on March 4, 2009 - 12:17 am
@Brian Utterback Re: are there any drummers that use a click track but use their own judgment and deviate from it?
My brother does studio drum work. He told me of a story where a sound engineer had him re-record a few bars where he fumbled a stick. The engineer didn’t/couldn’t get the drums muted from the playback in the headphones during the retakes and my brother is too humble to say something. It took a ridiculous number of retakes because it’s nearly impossible to tune out something like that. My point is that if a drummer where capable of making such deviations by choice, they are certainly not using a click track. Maybe you could concoct an inconsistent click track, but that is probably going to cause more problems than it solves for similar reasons.
#94 by Void Main on March 4, 2009 - 1:01 am
Skip that request for the gnuplot source. It’s pretty straightforward. Thanks!
#95 by Matthew Boehm on March 4, 2009 - 2:34 am
If you were worried about performance, instead of calculating the running average by constantly recalculating the sum of the list with sum(list) you could update the sum by subtracting the popped value and adding the value passed into the function each time.
I coded this up as “Method 2” and got the following results:
(1 million items)
Method 1 time: 1.21409173512
Method 2 time: 0.618820199215
(10 million items)
Method 1 time: 12.1816600993
Method 2 time: 6.23038151497
I actually expected the two functions to have a different big O instead of always running in half the time, but then remembered that the sum() function is being called on a list that isn’t growing in size past 16. Increasing the number of samples in the running average (max) does make method 2 increasingly efficient, but in the real scheme of things, finding running averages for 1 million beats in about a second is probably good enough.
#96 by Chris on March 4, 2009 - 6:09 am
Interesting article. I’d like to see a similar one done on detecting autotune, but I know it’s a significantly more difficult task to extract a pitch in a sea of frequencies. Though anyone who’s watched Melodyne DNA at work will know it’s probably not impossible.
#97 by Dave on March 4, 2009 - 8:48 am
Fun stuff! I’m analyzing a bunch of stuff I recorded, and I’m finding I pretty consistently play ahead of the beat. Thanks for the article!
#98 by MMI on March 4, 2009 - 10:28 am
The dialogue afterwards… well, I’m reminded of the story of the blind men examining the elephant. Each of them concluding that they were examining something other than an elephant.
Much of the discussion has focused on drums and the regularity of the player in question (if indeed there is one). The issue of timing in music is much more complex than the regularity of tempo as expressed by any single instrument. Another important bit is the synchronization requirements of various players. The “rhythm section” needs to be “tight”. Click tracks were a hack to help solve the problems raised by multi-track recording (when different parts were recorded separately, perhaps at different times and places). This hack is still with us.
Computers in the studio are a game changer. However, it’s all new enough that there is no “standard” way of working just as there isn’t a consistent level of skill and experience with the new tools.
The thing that is certain, to me, is that programmers will continue to be interested in music (and will dabble). Musicians will continue to look for new sonic frontiers (and some will find us). And the overlap between these two groups will continue to fascinate and entertain. Awesome.
#99 by Claude on March 4, 2009 - 10:59 am
Doesn’t Pro Tools have some sort of auto correct function for the timing fluctuations?
#100 by Big Time Producer on March 4, 2009 - 11:25 am
What a funny test this is considering a few things.
We all see Drummers wearing headphones that we assume are playing to a click either live or in the studio. Well just because they wear them not mean they are playing to time.
I would say that 99% of drummers cannot play to a click in any way shape or form, and the ones that can, do not exactly play like Tony Williams if you get my drift.
As people have noted the issue comes with trying to put everyone else in time, the editing is hell.
Also please note that when most drummers are playing their hi hat and snare at the same time…they flam. The biggest nightmare of course is the Kick drum, which for me most drummers cannot play 4 bars in a row in time.
Porcaro and Gadd are 2 in a million. Now Lars from Mettalica? That is what happens when either someone cuts the 2 inch tape a thousand times or someone just puts samples on crap playing.
There is a reason there has been a few Bass players…his time is brutal.
Biggest problem with drummers for me is they have too many drums to hit. Give them a kick snare and hat, then redo a click test.
#101 by peter on March 4, 2009 - 11:35 am
I would be curious to see some analysis on jazz drummers of the past. some of the big band drummers were famous for being able to keep apparently ‘perfect’ time. Buddy Rich in particular.
btw, check earlier metallica recordings, according to interviews around the time of ‘and justice for all…’ metallica used click tracks a lot as they recorded the guitar first and they had a lot of very specific tempo changes that were programmed in.
#102 by Creighton on March 4, 2009 - 12:27 pm
You can program click tracks to have a change in BPM per sections of the same song.
#103 by Ron on March 4, 2009 - 4:31 pm
As a drummer with more than 30 years of experience, I can see the value of a click & often practice with one in my ear, however, I wouldn’t use one for either recording or playing live. If you’ve got a good enough ear and a good sense of time, you shouldn’t need one. In addition, drumming in a live situation is really partly about keeping time and partly about getting into the “pocket” and really feeling the song. For me, having a click in a live situation or recording session would be too distracting.
Regarding the charting of the songs, it would be interesting to see what some of the great studio drummers have done (i.e. guys like Steve Gadd on “Fifty Ways to leave your lover”).
#104 by Big Time Producer on March 4, 2009 - 6:14 pm
Just listened to Breaking Benjamin Firefly as posted here.
I guess you should teach your click counter to count bars, there is a bar of 2 in the bridge, which basically means that any song you have analyzed with any odd amount of bars will not show an even line.
This would make your comparison extremely inaccurate no?
#105 by Dub on March 4, 2009 - 6:22 pm
It would be *really* interesting to see a graph of the first four minutes of so of Dave Brubeck’s 1959 classic “Take Five”. (In fact, it may be that the bass line is more consistent rhythmically than the drum line in that piece…) I think it would look *nearly* like a click track, and these are jazz musicians, so it’s a pretty solid bet to be click-track-free.
#106 by Jesse Cannon on March 4, 2009 - 6:45 pm
I find this story to be highly flawed and incorrect. I have posted a rebuttal at:
#107 by plamere on March 4, 2009 - 7:25 pm
Great rebuttal, thanks. I’ve learned a whole lot about drumming, and audio engineering by reading all of the responses to my post about searching for click tracks. I’ve certainly learned that a click track is not just a metronome that mindlessly keeps a tempo, instead it is something that can be pushed and prodded in all sorts of ways. So, when we see a curving line on one of my click track plots that may not necessarily mean that no click track is involved, it could mean that a more sophisticated, ramping click track or some more advanced audio editing is involved, using the techniques that you suggest. However, I do suggest that when the tempo deviation plots show no deviation – essentially a flat line (like with the britney spears, nickleback, breaking benjamin and hungry fathers tracks) then most assuredly we are seeing a track where the master clock was a machine and not a human.
#108 by Nick on March 4, 2009 - 11:58 pm
You’re wrong, actually. The beatles used a click often, most
notably on All You Need Is Love.
#109 by wheatus on March 5, 2009 - 2:15 am
In my life, to click or not to click is more a measure of the song and what it needs….Very rarely can you save a track (or a shitty drummer) with a click. If the song has some sort of backing track or loop in it then a click is hard to avoid…but our drummer Kevin Garcia can pretty much do the whole, ‘feels like a click without one’ thing, OR NOT if needed and there are other pros who can do that too…..This I know from comping his takes. Pro Tools quantizing is more likely what you are measuring in these graphs and not a the presence of a click track.
But the idea that a click makes for a weak drummer is nonsense.
The Police for instance….I cannot imagine what their studio lives were like….Copeland is amazing, and ALL OVER THE PLACE….Graph an old Police track and see.
But put him on a click and those tracks would suck.
And then there’s a guy like Steve Ferrone (Tom Petty and The Heartbrekers)…He’s a swiss clock…or Phil Rudd, who is both a clock and a meat plow. And the music demands it in these cases….Copeland playing for AC/DC would be a disaster.
A click track is a studio tool…to make a great album you need a great drummer one who can swing and click and all points in between. Instead of focusing on who does and doesn’t play with a click you should be worried about which producers chop the life out of their takes with Pro tools and which do not….now that’d be a graph for the ages.
brendan b brown
#110 by dejanilly on March 5, 2009 - 3:42 am
hm,….i saw the interview with Lars Ulrich and he said that he DIDN’t use the click track only on the first album (kill ’em all) and on the last album (death magnetic)…..so for sandman ?
any thought on this ?
#111 by plamere on March 5, 2009 - 7:58 am
dejanilly: As many commenters have suggested, it is possible to have a click track that varies in speed over the course of the song, so perhaps that is what is happening here.
#112 by droon on March 5, 2009 - 8:15 am
This is cool..
here’s an idea:
have your drummer/band record the song without the click track, then, “straignten” it, do your anaysis as shown here, but then stretch and unstretch the recording so it fits a straight beat, then you can do you protools magic, when the whole track is mixed down and re-assembled, have the band play it live again, and use that recording to stretch the timing back to it’s original dynamic!
#113 by ffualo on March 5, 2009 - 12:16 pm
I wonder if disturbances can be picked up in African and jazz percussion…
#114 by JT on March 5, 2009 - 1:21 pm
It is definitely possible for a drummer to produce a strikingly flat line without using a click track. At a music industry convention in Los Angeles last year a “most accurate drummer” contest was held, where drummers were invited to play a particular rhythmic pattern on a pad attached to a computer. The computer maintained an internal click track, and measured the drummer’s timing against the computer’s clock. The drummer would play the pattern for 60 seconds and then receive a score. The most accurate drummer would win.
As expected, most people had reasonable variety in their timing accuracy, but a few were remarkably consistent. The winner of the contest, a guy named Erik Truelove, had a score of 758 out of 800 – that’s 95% accurate, and if graphed above, would be probably be in the “uses a click track” category.
#115 by Whiskeytown on March 5, 2009 - 1:35 pm
I liked foo’s comment above and your slap-dash approach to the scientific method. It reminds me of some of my favorite episodes of Mythbusters i.e accurate enough to get your result with a million other experiments waiting in the background.
Looking forward to an update or someone else going PhD on this shit.
PS. I bet ‘radio lab’ or ‘sound opinions’ would be interested in this.
#116 by James on March 5, 2009 - 1:35 pm
It’s extremely unlikely that someone would record to a metronome with a varying tempo. There’s a ton of reasons why you wouldn’t record that way.
First off, if you don’t care about a steady beat (and you should think hard about that), then the drummer should be the one varying the tempo, and they should do it live, not to some pre-recorded changing click track. Now, the reason the tempo varies in this kind of music is the band members are following the excitement of the song. But it’s hard to follow the excitement of the song when you’re merely turning a tiny tempo dial with your finger, or, even worse, when the recording engineer sets it all up or is changing it live.
It’s much easier for the drummer to vary the tempo live than to play to a constantly changing click track–playing to a perfectly steady click track can be a challenge enough. It’s almost ridiculous to even suggest it. Perfectly matching a pre-calculated curve? Way harder. Perfectly matching pre-recorded changes made by a human? Even harder than that. And that’s even if you make the changes yourself.
As for playing to a click track being a bad thing, there’s no orchestra in the world that would play a symphony without a conductor. The only reason it’s thought of as “bad” is because bringing in a click-track is an insult to the skill of the drummer and his/her inability to keep a perfectly steady tempo, and not necessarily because it can potentially ruin the feel of the song. 90% of people would not even consciously notice whether the tempo is perfectly steady or not (hence the need for programs like yours). Generally, though, the audience unconsciously prefers a steady beat with large volume (dynamic) range. People get bored with a song with an ever increasing tempo that is constantly at full volume (most amateur bands and a lot of pro bands naturally do this and have to work against this tendency). The tempo changes wear the audience out, and the unchanging volume is simply boring.
#117 by Christopher on March 5, 2009 - 2:37 pm
Suprised about weezer? I don’t really dig the new stuff, but the first two albums and random tracks through out the later career are good and Pat Wilson is a very talented drummer, both stylistically and technically. Sure it’s just rythem back beat a chunk of the time, but that’s what this kind of music calls for. Check the fills in “say it ain’t so”, in the last chorus there is a fill that is almost hidden and yet amazing in it’s instrumentation, location and implementation. Style and technique. Just because you don’t like someone (based on your affection for Breaking Benjamin, i can certainly see why they aren’t your cup of tea) doesn’t mean that they don’t have ability.
I would have figured that spears was drum machines.
#118 by plamere on March 5, 2009 - 2:40 pm
Christopher – I actually like Weezer quite a bit. I think I have all of their albums and have managed to get my teenage daughter to like El Scorcho! My surprise is mainly comes from Weezer being an established band with musicians that may not get into the studio at the same time very often, so I would expect that a click track would help them collaborate better.
#119 by Jim on March 5, 2009 - 3:13 pm
I wonder what Rush would look like on that list. Neil Pert is often considered one of the best drummers in the world but it wouldn’t surprise me that he used a click track. He is to meticulous with his work.
#120 by plamere on March 5, 2009 - 3:39 pm
Ok … so I have had more requests for me to analyze Rush than just about any other band, so suggest a track or two that would be good (perferably a track that doesn’t have any explicit tempo changes) and I’ll run the analysis.
#121 by Jonny on March 5, 2009 - 4:02 pm
Very very interesting. I use all sequenced drums and I haven’t done a lot of tempo variation (other than tempo speed ups) but I am thinking of looking into a bit just to see if I can add a bit of “life” to my tracks.
#122 by geekamongus on March 5, 2009 - 4:11 pm
As a drummer of 26 years, I must say that I despise click tracks. They are to drummers what the Auto Tune is to singers. http://en.wikipedia.org/wiki/Auto-Tune
I was in an awesome rock band that I loved playing with. We hit the studio to start on an album, and the lead guy wanted me to use a click track to make it easier for the other guys in the band to add their parts later.
I attempted this, and did a pretty good job of playing along with it. However, he would sit there and listen to it over and over, noting “the bass drum is slightly off”. This was so minutely indistinguishable it was unreal. From that point on, things spiraled downward, and they kicked me out of the band the same day I was planning on quitting.
Anyway…just some random notes on me and my life, for which I’m sure no one will care about ;)
Thanks for the article…it was fascinating.
#123 by grooveiron on March 5, 2009 - 4:19 pm
Where do people get the idea that a steady tempo is a mark of a superior musician? Most pros know that it’s the details of timing which separate the good from the amazing. True for many kinds of music (not Trance!)
On the one hand people are ready to trash drummers who can’t keep a steady beat, on the other they trash drummers who use a click track???
Neil Peart of Rush is a good example of an accurate and soulless drummer.
Clyde Stubblefield (James Brown’s “Funky Drummer”) is crazy inaccurate.
Who gets sampled more?
#124 by ambignostic on March 5, 2009 - 4:40 pm
Case in point: The Grateful Dead.
On that note, it might be especially interesting to take two versions — the studio version and a concert version — of the same song by the same band, and juxtapose their beat duration plots.
#125 by Resonant Serpent on March 5, 2009 - 4:43 pm
It’s really a moot point to try and see if someone is using a click track. There is so much post production that goes on after the drums are recorded in modern music that you would have to have the original recording to really see if there was a difference.
You could do a torrent search for ‘stems’ and get some of the original tracks to older songs and the original drum recordings of Led Zeppelin, Queen, Nirvana, etc.
The flat lines your getting on drumming have more to do with Beat Detective and Drumagog.
#126 by Torrefaction on March 5, 2009 - 4:45 pm
Hey, I’d love to see the analysis of Nine Inch Nails. Preferably one track from something off of Year Zero, and another track off of With Teeth.
Mostly this is curiosity of two things.
1.)The timings Reznor uses throughout his songs vary on Year Zero, and use non-traditional rhythm so it could be cool.
2.)With Teeth has Dave Grohl on live drums, and I’m curious if he used a click track (Not all tracks are with Grohl, so I’d appreciate if you made sure he was on the track you use.)
#127 by Gingo on March 5, 2009 - 4:53 pm
Lars from Metallica has been using a Click track for a very long time. He has said so publicly. Sandman was recorded on a click track.
#128 by Guitar Bob on March 5, 2009 - 4:54 pm
True, but a groove is a very fun thing to play over.
This Monday night my drummer had a real tempo ‘burble’ towards the end of the 1st tune, wandering up and down. I guess wasn’t warmed up yet or something. It was not only embarassing, but just flat no fun.
#129 by andy on March 5, 2009 - 4:55 pm
This is a pretty bunk analysis for a lot of reasons.
#130 by Jason Bean on March 5, 2009 - 5:25 pm
Wow, that is indeed some cool stuff dude!
#131 by Anthony Pittarelli on March 5, 2009 - 5:31 pm
wow that one the main reasons i hate todays music. man, zeppelin and the beatles kicks ass….. metalica used to.
#132 by Mark on January 25, 2010 - 12:49 am
is this The Anthony Pittarelli the kick a** drummer of the band Child?
#133 by Jonathan on March 5, 2009 - 6:34 pm
It’s not very likely that this has anything really to do with click tracks.
The reason for 90% of the flat plots: Protools quantizing (or other equivalent manipulation). A time line will reveal it: highly produced bands AFTER about 2000 will be flat, those before will be a lot less flat.
#134 by Glen on March 5, 2009 - 8:11 pm
Yesterday I made a post about how much I love Moonlight Sonata by Beethoven. Then I started thinking about what really made it cool and it was the notion that the timing was part of how he expressed the song. No doubt if you played it quantized midi it wouldn’t have the same feeling.
I wonder if your graph detects the use of “swing” emulation and other programmatic timing variances.
Keep up the research!
#135 by Manny Ortiz on March 5, 2009 - 10:30 pm
The Police used a click track on their last tour…
Sting was always pissed at Stewart Copeland for his tempo changes.
I have upped a sample mp3 with the click track that is recorded from Sting’s Ear monitor.
#136 by Rene on March 5, 2009 - 10:36 pm
I know a few drummers who have such steady time that you would think there is a click… in fact, with one of them, when there is a click he can be so accurate that the sound of the click gets hidden behind the drums and you can’t even hear it when recording with him.
Also, sometimes drummers prefer to “drift” around the click, pulling and pushing… I guess you would see hills and valleys, but around the same axis.
In addition when program loops and whatnot (instead of a click, to make it more interesting and inspiring) people will map out tempo shifts… up a little at the bridge… yada. This would probably look pretty programmed.,
On top of that, so much music today is protooled within an inch of it’s life, every beat fixed, so who knows what actually happened. It’s like the corny protools joke: the engineer says after the first take “OK that sucked, we’ve got what we need” and start editing.
why not make the tempo axis reversed, so increases are a line rising.
Ok… not sure why I felt the need to go into so much detail on this random blog… back to work.
#137 by business on March 5, 2009 - 11:05 pm
I wonder how classical musicians would measure up.
#138 by Endif on March 5, 2009 - 11:40 pm
Clicktrack is not the most plausible explanation unless you’re dealing with live performance: most CDs are recorded digitally. Excessive variation in percussion can and is usually quantized or edited out in the studio.
#139 by Shaun on March 6, 2009 - 12:33 am
Can you try out a couple 311 songs? Chad Sexton is known for being able to keep amazing time, and I know he doesn’t wear headphones during live sets. Maybe try a couple live songs? I have their live album if you want a rip.
#140 by Pat Tufts on March 6, 2009 - 12:40 am
I’d be interested in how Frank Zappa’s drummers come out, particularly Vinnie Colaiuta.
Zappa was known for wanting his musicians to stick to the score exactly, and Colaiuta had an ability to play ‘impossible’ parts on the drums without any sense of strain.
Excerpt from the above link:
‘He’s one of the most amazing sight-readers that ever existed on the instrument. One day we were in a Frank rehearsal, this was early ’80s, and Frank brought in this piece of music called “Mo ‘N Herb’s Vacation.” Just unbelievably complex….’
If click tracks existed at that point, Zappa would very likely have required them. It would be interesting to see a Zappa drummer’s performance from the pre-click-track and click-track eras.
#141 by gd on March 6, 2009 - 1:33 am
now, itd be fun to try to find which drummer most closely resembles a steady click track without actually using one.
#142 by Maurice Atkinson on March 6, 2009 - 4:23 am
Click tracks??—–Earl Palmer. ‘Nuff said.
#143 by Jop on March 6, 2009 - 5:59 am
Good work, I like it. Interesting to read about clicktracks and what it does to one’s playing. Personally I’m a big clicktrack fan, especialy when you work in the studio, because it makes the drums that much easier to work with. I also find that a lot of premade backingtracks/loops are click related.
I’m quitte into the whole click thing at the minute, and where drummers place their notes. If you want to look into something this might be cool. Some drummers seem to push or ‘relax’ the beat, although still playing in time. Some drummers play certain notes early or late, to get a certain feel in the music.
anyway, good work, I like it!
#144 by dwight krueger on March 6, 2009 - 9:28 am
hate click tracks, try to avoid using them at any time. would any of the great music from the 50s,60s, etc… been any better with a click track. i think not, maybe the opposite.
#145 by Matt B on March 6, 2009 - 9:41 am
I’d love to see how my band’s songs stack up – I’m the drummer. I can email them, or you can grab them from http://www.myspace.com/theslackwaternews.
#146 by Christopher Small on March 6, 2009 - 9:58 am
Hey, Paul. I’ve heard that a lot of what U2 does live is synched with sequencer tracks, but the drummer (Larry Mullen Jr.) hates wearing headphones, so a click track into his ears won’t work. Instead, they mix a time-synchronized maraca sound into his monitors and he lines up his drumming with that.
I wonder if this would result in a plot that would still look pretty flat, but with more variance because of the other audio stimulus he’s getting.
#147 by putty on March 6, 2009 - 10:48 am
There seems to be an impression here that drummers either always (or at least usually) use a click, or never use a click track. In my experiance as a drummer, it was somwhere between those two. I never used a click live. But sometimes we used a click track while in the studio if someone thought it was needed, otherwise it was usually left out. If a tune was going to have tempo changes, it would be far too expensive and time consuming (at $400/hr studio rates) to program a midi click with the tempo changes since we could just do it by feel in a few takes anyways. Some songs might actually use a click for parts (usually at the begining, or to keep time when only guitars and vocals were playing) and then stop using it later on.
Using a click was never a problem for me because I often practised with an amplified metronome and was comfortable playing that way.
Also, there are at least two distinct ways a drummer can deviate.
1) Tempo – the tendancy to speed up or slow down throughout the song (or as someone else mentioned during fills etc).
2) Accuracy. It often happens that a drummer may produce a note that is “off” while actually remaining in tempo with regard to the rest of the notes they play. Picture yourself laying a tamborine overdub on a song that’s already mostly finished. Tempo will not be an issue as it’s already set by the bedtracks, but you will be very aware of your accuracy (and the drummer’s ) as you try to match the drummer’s notes with your tamborine hits in the overdub.
Sometimes a click track may be used to make it easier to detect accuracy errors as in #2 above than to keep the tempo steady.
Finally – it should be possible to use regression analysis (statistics) to formulate a more scientifically acceptable method to determine whether a particular track was recorded with a click or not, as opposed to just eyeballing the graphs.
#148 by Poison3k on March 6, 2009 - 11:07 am
click track? just hit those damn drums hard and enjoy the song!
#149 by putty on March 6, 2009 - 11:08 am
One more thought. For tracks where you determine a click was not used, it would be interesting to compare the standard deviations in relation to the tempo itself. It’s entirely possible that some drummers keep tempo better at certain bmp rates. Personally, I’ve always found slower tempos more difficult to keep.
So you could compare the standard deviations in graphs for Led Zepplin’s “Dazed and confused” to “Whole lotta Love” and try to figure out if Bonham was kept better time at faster or slower tempos.
#150 by Mark on March 6, 2009 - 12:50 pm
As a drummer (and bass player), I have been involved in recording over 200 songs since 1987. In all but a very few sessions, I used a click. I have also used sequenced backing tracks in live performances. I don’t find these results at all surprising. What I WOULD find interesting is songs that SOUND like they were actually played by a drummer (on a non-electronic) drumset, but are actually a collection of sampled percussion sounds, locked in and quantized to sound “real”.
#151 by Ben on March 6, 2009 - 2:08 pm
#152 by greg on March 6, 2009 - 3:32 pm
surprised about weezer? have you listened to pinkerton lately? the speed is all over the place – which in my opinion, adds a cool lo-fi/immediacy to the songs.
as a former drummer, i always had a hell of a time playing with a click. it is a skill unto it’s own.
#153 by Greg Alexander on March 6, 2009 - 3:58 pm
Well I have advice for a better clicktrack detection technique. There is an old (analog) tuner technology called the stroboscope. The idea is that you have a spinning wheel in front of a light, and the light is driven simply by the audio waveform (maybe slightly filtered). The wheel is designed such that some portions are white and some are black, and it spins at the frequency that you are testing for. If the audio signal contains any frequency components that are in sync with the spinning wheel then this will be perceived by a viewer as the wheel standing still. If it has components very close to the spinning wheel’s rate, the wheel will appear to turn one way or the other. It makes a very good tuning device because an error of only 0.01Hz can show up as the wheel spinning at 6rpm, which is still very discernable visually even while 0.01Hz is not very discernable to the ear.
This process can work at very low frequencies (i.e., the range of 1Hz to 0.25Hz at which rhythm is perceived). The analysis is _much_ simpler than Echo Nest. And the neatest thing is that it shows cumulative errors as well as instantaneous errors. For example, a drummer even with a click track may choose to play one measure fast and then the next measure slow (a phenomenon which is evident in some of these plots), making it difficult to be certain he is using a click track. But if, after the deviation, he again returns to the absolute tempo (i.e., in a 60bpm song, if beat 200 occurs at second 200, no matter what variations in between), you can be quite certain he is using a click track. This shows up very clearly in the stroboscopic view.
A simple implementation of a stroboscope on a computer is to do a scanline graph where the brightness of each pixel is dependent on the amplitude of the corresponding sample (though you’ll need a low-pass filter such as averaging to do very low-frequency analysis)…Start at the upper left, proceeding to the right, and when you reach the reference period (i.e., 1 second for 60bpm), then proceed to the next scanline (restart at the left side of the screen). If there are vertical lines visible in the resulting picture, the signal and reference are in sync…if there are diagonal lines, it is a little off one way or the other, and if it looks like gibberish then there is very little correlation.
I consider the stroboscope-style analysis to be superior because a good drummer may return to exactly the same tempo that he started but will almost never return to the same absolute beat (he may return to 60bpm but it is unlikely that beat 200 occurs at second 200).
Just a thought.
#154 by plamere on March 6, 2009 - 4:00 pm
Interesting idea, I may try this if I have a few spare minutes!
#155 by Jeremy Wilms on March 6, 2009 - 4:45 pm
Firstly, it would help me if the vertical measurement should be BPM. It might make the relative variations clearer to the average musician.
Secondly, I am creating a machine that determines what kind of people need a machine to tell them who uses a machine (other than their instrument which is a machine) to play music (which in itself is also a machine)(as are the people who are playing the music).
#156 by plamere on March 9, 2009 - 7:16 am
Jeremy – BPM scale … good idea. As for the meta-machine thing .. I think you might be making fun of me ;)
#157 by Grumdrig on March 6, 2009 - 5:00 pm
Very cool. In return I rewrote the code more compactly, and looplessly. I didn’t try it on real data because I don’t have any.
return float(sum(s)) / len(s)
audiofile = audio.LocalAudioFile(inputFile)
beats = [beat.duration for beat in audiofile.analysis.beats]
times = [sum(beats[:n]) for n in range(len(beats))]
WINDOW = 16
avgs = [avg(beats[max(n-WINDOW,0):n+1]) for n in range(len(beats))]
base = avg(avgs)
output = zip(times, [a-base for a in avgs])
print “\n”.join([” “.join(map(str,o)) for o in output])
#158 by plamere on March 6, 2009 - 5:50 pm
Cool – no loops – so I’m new to Python, is this loopless approach the Python style or more of a parlor trick? To me as an old Java programmer, the version with loops reads easier – does this version read easier to you as am experienced Python programmer? If I continue to generate Python examples for pedantic purposes, is this style prefered?
#159 by Pascal on March 6, 2009 - 7:14 pm
Very interesting stuff. I want to take your code out for a spin on some other tracks, but I’m a little concerned by The Echo Nest terms of service:
Licensee further grants to The Echo Nest an irrevocable, non-exclusive license to any and all data provided by Licensee to The Echo Nest in connection with Licensee’s use of the APIs, including without limitation, any user data, text relating to Licensee’s music catalog and/or website and any sound recordings, musical compositions or other copyrighted works. Licensee represents and warrants that all such data provided by Licensee to The Echo Nest shall not violate any third party rights.
Isn’t that a little, ah, overly broad? *Irrevocable* license? To my Web site contents? To any and all music I submit to the API?
Also, as a scholar, I might be working with music to which I do not own the rights but can still use as pat of “research and study” copyright exceptions. How do you handle that? In fact you yourself obviously do not own the copyright to the works analyzed in this post… Wouldn’t that mean you are violating the terms of service?
#160 by plamere on March 6, 2009 - 7:47 pm
Pascal – thanks for the comments and expressing your concern. Indeed, we know that the current terms-of-service is overly broad, and we are in the process of redrafting them to make them much more developer friendly. The new terms should hopefully be drafted and posted soon. If you want to get started before the new TOS are published and are concerned about a particular provision, send me an email – I’m sure we can quickly work things out so that you can do what you want to do without worry that we are going to take your website, your house, your car and your computer.
#161 by Grumdrig on March 6, 2009 - 7:30 pm
This style (using list comprehension) is newer and jazzier. I certainly prefer it when it makes sense, and in my opinion is often more readable once it’s at all familiar. I’m not sure it makes sense here, frankly; it at least looks inefficient. E.g.
[sum(beats[:n]) for n in range(len(beats))]
is O(N^2), whereas your calculation is O(N). The computation of avgs is also a little awkward; this idiom works best when its a transformation on each element individually. I went ahead with it to show off some python, and because refactoring the code that way is kind of fun. :)
#162 by michael on March 6, 2009 - 8:46 pm
At first glance I was intrigued by this, presumable scientific study of ,the consistency in the meter of drummers. Upon further analysis, I discovered too many inconsistencies with this hypothesis. First,with all the technology in recording programs, even the sloppiest of drummers can sound metronomically precise. Also,there are graph screens in audio recording programs that can show( in recorded and/or real time) how consistent a drummers beat is to the 120th of a beat.2nd,an experienced drummer knows how to play around with the time( playing in front of,right on the beat or behind the click).There are drummers that simply just cannot play to a click and would vary in tempo anyway. To the contrary there are drummers well in tuned with time and play very consistently without the aid of a time keeping device.
One question I have is, does his procedure evaluate time change and meter changes? However, I do agree that there are complete differences ,to a musical piece, when comparing the natural feel to a drummer and the more precise and accurate recording with a metronome.Of course then again, this too can be altered by programs. Music is for enjoyment not analyses.
#163 by tanveer on March 7, 2009 - 12:57 am
I really can not understand it.
#164 by FS on March 7, 2009 - 1:48 am
Why not run this in a loop through a whole MP3 collection and post the results on a site? That would be cool, since you will have a statistic to correlate the assumption, and pinpoint where in time people started to use click tracks. Just remember to tag the MP3 collection for date of release…
#165 by plamere on March 9, 2009 - 7:14 am
Thanks, good idea … if only there were more time in a day!
#166 by j m on March 7, 2009 - 2:05 am
Where’s the r^2? I don’t get it.
#167 by evan on March 7, 2009 - 7:52 am
#168 by Arrenlex on March 7, 2009 - 4:00 pm
This is really cool. I’ve signed up for a dev key and have been playing with it.
Muse’s Knights of Cydonia makes the program crash! :(
em@sam:/tmp$ python click.py Muse\ -\ Knights\ of\ Cydonia.mp3 > cydonia.csv
Traceback (most recent call last):
File “click.py”, line 31, in
File “click.py”, line 6, in main
beats = audiofile.analysis.beats
File “/usr/lib/python2.5/site-packages/echonest/audio.py”, line 130, in __getattribute__
value = getter(object.__getattribute__(self, ‘id’))
File “/usr/lib/python2.5/site-packages/echonest/web/analyze.py”, line 71, in get_beats
return util.apiFunctionPrototype( ‘get_beats’, id )
File “/usr/lib/python2.5/site-packages/echonest/web/util.py”, line 38, in apiFunctionPrototype
File “/usr/lib/python2.5/site-packages/echonest/web/util.py”, line 83, in parseXMLString
raise EchoNestAPIError(status_code, status_message)
echonest.web.util.EchoNestAPIError: Echo Nest API Error 8: Computation error (TRISTAN-0)
It’s a shame; I was really curious about that one, as the drums are spectacular.
#169 by plamere on March 7, 2009 - 6:33 pm
I tried it with my own local copy and got the same error. I’ll check with the back end guys to see what is up.
#170 by Charlie Morgan on March 7, 2009 - 9:56 pm
Thanks for taking the trouble to examine this in detail. I found the whole thing most interesting. However, I need to set things straight in a few areas, regarding your first paragraph.
I am a professional drummer, with over 35 years of experience in the Music Business, and I can tell you that playing to click tracks is not a new thing. I have been used to playing to clicks (with feel) since the late 1970’s. In fact, all the film soundtracks I have played on made extensive use of clicks, to synchronise the music to the film.
What IS relatively new to the industry is the use of drum machines, and computer-generated programmes that use high-quality samples to SOUND like a real drummer. For example, the first tune you tried this one was “Hit Me Baby” by Britney Spears, which I know does not have a real drummer playing on it! It IS a computer-generated machine.
I’d be interested to know what results you get when there IS a real drummer who knows how to play to clicks with feel. Try Elton John’s “A Word In Spanish”, for example, or maybe even Tracey Ullman’s “They Don’t Know About Us.” Or even “Open Arms” from Barry Manilow’s recent album, “Hits of the 80’s”
Let me know…
#171 by plamere on March 9, 2009 - 7:13 am
Charlie – thanks – and good points about click tracks vs. drum machines. Arren lex has created a click plot of Elton John’s ‘A word in spanish’. I’ve reposted it here: Look here: https://musicmachinery.com/2009/03/09/more-on-click-tracks/
#172 by elmegil on March 7, 2009 - 11:15 pm
You should check some Rush, given how Ghodly many people consider Neil Peart to be.
#173 by plamere on March 9, 2009 - 7:11 am
Look here: https://musicmachinery.com/2009/03/09/more-on-click-tracks/
#174 by Arren Lex on March 8, 2009 - 3:18 pm
Here is Elton John’s “A Word in Spanish”
#175 by plamere on March 9, 2009 - 7:09 am
Awesome plot – I’ve started to make a directory of follow on click plots here: https://musicmachinery.com/2009/03/09/more-on-click-tracks/
Yours is the first, thanks!
#176 by arrenlex on March 8, 2009 - 4:48 pm
This is Rush – The Enemy Within. It crashed trying to do 2112.
I have no idea what the two spikes at the end are though… maybe it got confused as the song fades out.
#177 by plamere on March 9, 2009 - 7:11 am
I think you are right. The raw analysis data includes a confidence with each beat. One could adapt the code to only output beat info for beats that are of high confidence. I think that would eliminate the spikes. Again, nice job!
#178 by Todd Totale on March 8, 2009 - 6:15 pm
I would love to see an analysis of one of AC/DC’s tracks featuring Phil Rudd, particularly something from Back In Black, Highway To Hell or For Those About To Rock. Rudd has always claimed never to have used a click track. Mutt Lange is a producer that (now) practically requires them. I want to know if Rudd is in fact telling the truth for those Lange produced AC/DC albums.
#179 by rich on March 8, 2009 - 11:36 pm
Very interesting. I can think of a lot of uses for this. Any chance of a utility for non-programmers to analyze favorite songs, our own playing?
#180 by plamere on March 9, 2009 - 6:44 am
I’ve been thinking about a web based app that would let you upload, analyze, graph and play with the click plots – I’ll post about it if I get the time to do it. Thanks!
#181 by Arren Lex on March 10, 2009 - 12:38 am
AC-DC – Highway to Hell, as requested by Todd Totale:
#182 by james on March 11, 2009 - 3:17 am
This is great, I always envisioned software like this, so glad to have found it.
I would love to see electronic songs analyzed. A great example would be to use Kraftwerk vs. some kind of new electronic techno song that has perfect rhythm and tempo from a digital sequencer.
The interesting thing about Kraftwerk is that they were using analog sequencers which weaved in and out of time very subtle (depending on voltage). Digital technology prints out your song to the correct beat and rhythm within nanoseconds.
Autechre tracks would be great too because they are known for editing slight tempo fluctuations to create abstract grooves and rhythms.
Have you tried songs with strange time signatures, other than standard 4/4? How does it react?
What about running an audio recording that doesnt have any rhythm? Like a news broadcast with various sound effects. I wonder to what sensitivity the program is detecting rhythm.
What about placing two completely different songs over each other? How would the software react? How would it ‘pick and choose’ or use all of the rhythms to determine the deviation?
#183 by Sammy Samcore on March 11, 2009 - 8:40 am
I love the analysis and am somewhat interested in learning Python. Great job! By the way one of the greatest drummers in the world – Aaron Gillespie of UnderOath uses a click track on stage and in recording as well.
#184 by ifly2gethi on March 13, 2009 - 10:14 am
Well done, bravo zulu, great article proving the science behind the madness which Elwood Blues (aka Dan Aykroyd) predicted in the early 1980’s – “You know, so much of the music you hear today is preprogrammed electronic disco, you never get a chance to hear master bluesmen practicing their craft any more. By the year 2000 and 6, the music known as the blues will exist only in the classical records department of your local public library…”
In any case, a click track, chickapuckka-encrusted, band-in-a-box will NEVER be able to emulate the one thing that musicians have, who are actually adept at operating an instrument: FEEL.
Again, well done.
DEATH TO GUITAR HERO (I leesen to good Soviet Symphony – Lynyrd Skynyrd!!)
#185 by Bjorn on March 13, 2009 - 12:20 pm
I’ve recorded and edited a lot of classical music, and the stuff done by pros is a lot easier to splice together: even if recorded weeks apart in different rooms they can match the tempo and pitch exactly, but amateurs can never match that stuff from minute to minute.
Recently, modern music has started using click tracks so that stuff can always be moved around and edited at whim, and the music tech blog the music machinery wrote an interesting piece where they actually analyzed the variations in tempo throughout the songs complete with charts and graphs. Of course, theoretically, people could have been using metronomes long ago and tight drummers could be keeping very consistent tempos without a click track today (or preprogramming variable click-tracks), so it’s not hard science, but it’s an interesting read nevertheless. I know I’ve heard more than a few tracks where I thought “man that’s tight!” and there was probably no click track involved.
Personally I don’t see anything wrong with using the tempo “click” track, even if the implication is that it’s a crutch. A lot of great music has been made possible that would never have been possible before because of the ability to drag and drop music in a grid. At the same time, I think the if you don’t need it, don’t use it philosophy is probably a good one.
For his next exploration, I hope he looks at tracking the pitch and tempo of a college a-capella group. NIGHT-mare!
Thanks to http://www.extrapepperoni.com for this one.
#186 by MeanOnSunday on March 13, 2009 - 5:07 pm
Neal Peart definitely uses a click track when playing live because of the sequenced synth parts. You will see him put on headphones for those tracks. This same reason is part of the legendary Keith Moon, Cow Palace story. Forced to use a click track on the Quadrophenia tour Moon was so stressed that he overdosed on tranquilizers and had to be replaced by an audience member.
#187 by Peter Andersson on March 15, 2009 - 1:16 pm
Here’s a way of speeding up the process by analyzing whole albums at a time; get a BPM counter (I use the free one from MixMeister, at the bottom of this page: http://www.mixmeister.com/download_freestuff.html) and watch the digits, if EVERY song on an album gets to have .00 (or .99/.01 as it has a margin of error of .01) then you can be sure the drum tracks were all laid down by a drum machine. There’s a 1/100 chance that a random track gets the digits .00, there’s a 1/10000 that two tracks in a row get it, there’s a 1/1000000 that three in a row gets it, and so forth up to “astronomical chance” for an entire album. If you don’t believe me compare an early album by The Donnas to their latest, or the #Ace of Spades” Motörhead album to any of their three latest, or if you’re not into that then the new Kelly Clarkson album (although I don’t have her previous stuff).
The reason I discovered this is something completely different, I’m runner who collects good songs with very straight drum tracks for leg coordination purposes, kind of a “click track” for my feet if you like (most running takes place within the 140-180 steps per minute area, although sprintwork can taky you up to 200-210). Check out my 986 songs long list here (below the top 50 list): http://k1chyd.adress.eksjo.com/steps-per-minute.htm
#188 by Issa Diao on March 26, 2009 - 11:32 pm
Actually, what your program is detecting is not the use of a click track as much as it is the use of time-editing software (for example Beat Detective). Different drummers play to click tracks very differently. Expert drummers can keep perfect time with or without a click. Proficient drummers can purposefully play around a click. And typical drummers, even when trying to play perfectly to a click, don’t manage.
From a recording perspective, the main purpose of playing to a click is not to move passages around (that can be done with or without a click). The actual purpose is so that you can run Beat Detective. What Beat Detective does, is single out the beats and sub-beats of the drum tracks and MOVE them into perfect alignment. Any drummer who can come close to matching the click can be made to sound perfect. I’m guessing that is what you are revealing with your program—equally interesting, I might add!
An extra point…Ray Charles’ finger snaps were featured in a Scientific American article a while back because their timing turned out to be EXTREMELY accurate. While this is rare, I have worked with a few drummers who have the same talent.
Personally, I like the feel that tempo drift can add to music. But Beat Detective is REALLY good at making average drummers sound much better than they are.
#189 by Paul on April 4, 2009 - 10:28 am
It’s an interesting discussion, about which I’ve still reached no real conclusion!
I’ve been to quite a lot of contemporary ‘classical’ concerts where some kind of click track was used, usually because of the use of prerecorded material, often on a laptop. However this can reflect an influence from dance/laptop electronica but also be married with free improvisation etc.
I would be curious to know how Jaki Liebzeit of Can measures under this system. Although he had a jazz background, in my understanding he intended to get away from that rhythmic freedom and was one of the innovators of what became known as ‘motorik’ rhythm (as was the band as a whole). I don’t think he used any kind of track.
#190 by Pat on April 16, 2009 - 4:01 pm
I’ve been recording drum tracks for 28 years and only recently began seriously working with a click track. My current project is an 11-song classic/power metal album with a whole lot of double bass patterns. In pre-production the takes were very lame with a board-up-the-rear feel and very few ad-libs. Like Grohl, discovering how far off my chops were from the click was mildly devastating to my ego, but I’m an old fart and it was all new material, so the band went back to the garage and worked it out with a metronome, agreeing on a tempo reference for each tune.
That was time well-spent. We finished tracking my drums last night, and the improvement between the pre-pro demo and the money performance is just stunning. Using the click forced me to evaluate my drum parts more from a listener’s perspective than a drummer’s, prompting me to scale down overly busy parts, simplify fills, and make other changes that were causing me and the band to work against the click track. We also opted to use electronic kick trigger pads instead of acoustic bass drums to eliminate that bleed into the overheads. This session was probably the hardest work I’ve done on the tubs in my whole career, but I’ve never sounded more like a pro than I will on this album.
#191 by Eric on April 17, 2009 - 12:14 pm
I’d like to see you graph some of the great studio drummers, like Steve Gadd or Benard Purdie on tracks known to be w/o click!
#192 by Danny on May 9, 2009 - 7:52 pm
The late Richard Tee (Steve Gadd’s long time collaborator) said that a professional musician (be it a drummer or anything else) should be able to play with a click track, regardless of whether they actually do so on a regular basis, and that Steve felt the same way. I tend to agree with him. Being ABLE to play with a click track means that your overall time is solid. If you want to go for a more natural sounding feel, then you can play without it, but the basic ability to remain in time should still be there.
Of course, Ringo, Bonham, and the J&M studio band (Fats Domino’s guys) never had a click track and look how great they turned out.
#193 by m.malloy on May 18, 2009 - 1:02 am
MUSIC BREATHES, THAT IS A NATURAL TEMPO WILL EXPAND AND CONTRACT, THE BEST PLAYERS CAN PLAY WITH TIME, THE FEELING THAT U GET FROM YOUR FAVORITE TUNE IS ALL ABOUT PLACING THE NOTES, IN THOSE TIGHT LITTLE PLACES OR NOT, NOTHEING WRONG WITH CLICK TRICKS AND NOTHING WRONG WITH RUBATO,
#194 by Joe Gilder on June 4, 2009 - 8:15 pm
This is so cool, and thought-provoking. I posted a link to it over at my blog.
I wonder if we “need” a click nowadays because a lot of drummers don’t learn to play with one? While there’s certainly the possibility to make things sound terribly robotic and drum machine-like, a click track can make a good group of musicians shine. There’s definitely a time and place to NOT use a click track as well.
Thanks for the post. I love your blog!
#195 by waymore on July 21, 2009 - 11:03 pm
any decent drummer will be very close,without a click.
for the audience it will sound great,the human ears
are only capable of so much.i never play with a click,
i do believe it takes the the feel from the music…
#196 by Andrew on August 5, 2009 - 11:41 am
My engineer frequently tricks bands who don’t like playing to click. He’ll let them do two rehearsals of the song and simply tap the beat into the computer and create a click track that follows their deviations. Then he’ll tell them that they’ll just have click at the top of the track as a reference and that he’ll turn it off once they’re into the song. Of course, all he does is turn it down slightly and then bring it back up. No client has ever caught him doing it, and it’s made his job easier.
As an arranger who usually has to demo the tracks in midi or samples beforehand, I can tell you that the use of click tracks makes my job much easier – it saves me an hour of work syncing BPMs in sibelius or Logic.
#197 by Take Solace on September 8, 2009 - 2:06 pm
Thanks for the info. Pretty cool idea. It didn’t answer for me the question of sterility, but I guess that’s probably unanswerable. Each song is different. And at least we know that Brittney Spears isn’t sterile..
#198 by Kyle on October 19, 2009 - 6:05 pm
Hey I was wondering if you could check some bands with faster drum beats like “As I Lay Dying – Forsaken” or “All That Remains – Empty Inside”.
#199 by Guitar Lessons For Children on November 21, 2009 - 5:25 am
Dude this was incredibly interesting. I don’t know if I’ve just been a little naive but I almost did not expect Breaking Benjamin to be using a click track.. although now that I think about it I guess I can believe it.
Anyway, this is super interesting, thanks for sharing your findings. You even gave the code which is awesome. I’m not sure if PHP is really made for this but I can still give it a go hehehe.
#200 by harvey on February 26, 2010 - 1:17 am
Awesome article! I’m just starting to get into drumming with a click track, gonna go plot stuff now.
#201 by thinksicily on March 3, 2010 - 9:37 am
An extreme example of accelerando in programmed music is “Alphabet Aerobics” by Blackalicious – check it out.
It’d definitely be interesting to check out more of the old jazz drummers. I’d be intrigued to see how folks like Big Sid Catlett and Cozy Cole compare with Art Blakey or Max Roach.
#202 by Mac Santiago on April 18, 2010 - 4:11 am
Extremely intriguing.Here’s my take on timekeeping and clicks.Sorry it’s a little winded.It’s part of an article I’m working on……A Poorly Played Note Well Placed…..”…..Is better than a well played note misplaced” This was my friend Wade’s reaction after I told him about the book that I had recently given birth to late last year entitled “Beyond the Metronome”.I thought to myself this probably was the best one yet e.g. it seemed that every musician I’ve spoke to regarding this subject had a reaction or experience in regards to tempo,groove,pocket or whatever one chooses to call that thing that gives rhythm it’s flow or value… the “Big Beat”.We can all agree is is the common thread through almost all music ,especially and not exclusive to that which has it’s rhythmic roots in Africa.i.e.folk,blues,jazz,swing ,rock,R&B ,hip-hop.,etc. ……Of course most players know that whatever one might think of anothers’ compositional abilities, choice of notes in an improvisational setting or even tone quality, judgments most readily fly in regards to a player’s intonation and time or inchronation(in-internal,chron-clock)…that is an individual’s own ability to create or carry that rhythmic objective both accurately and with an over-all sense of “steady”. This is not just for drummers and rhythm section players either.Everyone has an influence on the groove.A trumpet player playing a nice accent over an and -of- four hit…early,can bring down an otherwise solid groove.So, how does one go about getting this groove together.?Play with records? Play with a metronome?Sure ,if you think that following someone or something elses time is good enough.Getting ‘with it’ or in-sync is important but who’s tempo is it?The guy beating on an imaginary plane(the directors ictus) or some drummer or bassist reading a difficult passage in the music or focusing on the babe in the third row? And we all know how fun it is to practice with a metronome.e.g.that tug-o-war ,so often called a ‘crutch’ and generally the end result is unnatural at best.So, what do you do when you listen to a recording of yourself and you feel as if your in a race to beat one with a drummer who has sucked the life out of an otherwise great groove? Through many years of playing and discussing this somewhat vague topic I have come to the conclusion that in addition to being a good follower of the tempo,that is being able to ‘sync’ yourself to what’s going on, one should practice creating it and have a recognition of it just as one recognizes good intonation. Creating that steadiness that the music deserves should come from every player at any ability. In the practice room, a process of learning to subdivide and play with the metronome at quarter-notes is a great start but should be viewed as only the beginning. Gradually removing the metronome and replacing it with your own internal one seemed to be an appropriate step and one that was often missing in the aforementioned advice often given to students of the groove. With the exception of playing faster tempos to a slower click or displacing the click to beats 2 and 4(as the swing drummer’s hi-hat)an actual process has seemed to have ‘snuck’ under the radar or just plain taken for granted..As a result most young players and even some seasoned pros who can “play their asses off” can consistently show tendencies to rush through musical phrases or show an inability to recreate a given tempo from the beginning to the end of a tune,song or piece of music.In recording and even live, I’ve even noticed an increasing reliance on click tracks that lend themselves to players bouncing back and forth on either side of the beat in an effort to stay ‘with’ it. In the exercise below a student of the groove is given an opportunity to examine his /her own perspective of tempo.Granted the clicks indicated in the chart are metronomic(digital that is) and are to be used as a guide just as a metronome is but obviously diminished (halved) periodically.The idea is to take whatever you are playing, e.g.keep it simple and something you know real well.Remember well played is well placed…and as the click gets ‘longer’ in duration,substitute the click with your own estimation of where the groove is.Your own ‘internal click’.This way the metronome or click gradually becomes less of a guide for the time as your internal clock takes over.Ending up in perfect rhythmic unison with the the clicks is the ultimate goal, but how close you are to them (late or early)and it’s consistency is more important and should give you a fair estimate of your conception of the given tempo.Remember this: you must stay true to your perception of the groove.Becoming further away each time means you are not playing the given tempo or altering it to some degree or another i.e.rushing or dragging.At this point try again.Employ more subdivision this time.Maybe incorporating more non-essential movement as often players do.i.e. a slight movement of the head or shoulders. by not telling her/him what to play but where to put it..Well played”(in-tune etc)and …”Well placed”(rhythmically accurate).