The "Undone" Production diary

[soulata]

having listened to it a few times more:

+
I like the song
It's amazing what Melodyne does in right hands

-
Too many fills for me, groove is alright but fills spoil
it for me a bit, esp. the first one at 0:27

k

p.s.: Your sis really sounds like Bjork on certain parts (which is good in my book)

[Kim Lajoie]

Too many fills for me, groove is alright but fills spoil
it for me a bit,

That's an interesting point. I used to not include fills in my drum patterns, and I found that (verbatim) looped rhythms get really boring or really annoying.

When the looped rhythms are in the background, we tend to "learn" them and then ignore them. As listeners, we notice change. When the loop doesn't change, it becomes "invisible". Sometimes that's desirable - when the loop is meant to add texture or groove, but not stick out.

Sometimes it's not desirable. I consider the snare drum to be one of the most important instruments along with the vocal. I place it in the foreground. Looping without variation, a foreground instrument gets really annoying, really quickly. Especially something as cracking as a snare drum.

What I've mostly done is time the fills so that they occur in between or at the end of vocal phrases. This results in a kind of "call and response" between the vocal and the snare. This is particularly noticeable in the choruses, after the second and third vocal phrases. In fact, the space between the first and second phrases in the chorus are the only point when the rest of backing parts are clearly audible.

Of course, the snare is not the only instrument doing this. I have other instruments sounding in response to vocal phrases at various places in the song.

esp. the first one at 0:27

That, along with the fill at 0:38 are special cases - where the snare is playing with the vocal, instead of in response to it. I see it as a kind of modern tutti. It also has the kinda cool effect of disrupting the sense of rhythm - effectively stretching quavers (8th notes) to dotted quavers (3/16ths). This kicks off the verse with a tense instability before returning back to the relative stability of common time (4/4).

I'd be interested to know what you dislike about the fills.

-Kim.

[Kim Lajoie]

New notes about what happened when I played the final version to my sister.

-Kim.

[soulata]

esp. the first one at 0:27

That, along with the fill at 0:38 are special cases - where the snare is playing with the vocal, instead of in response to it. I see it as a kind of modern tutti. It also has the kinda cool effect of disrupting the sense of rhythm - effectively stretching quavers (8th notes) to dotted quavers (3/16ths). This kicks off the verse with a tense instability before returning back to the relative stability of common time (4/4).

I'd be interested to know what you dislike about the fills.

-Kim.

I know (or at least think I know) what you mean, but I guess what bothers me is this -

loop sounds like a real drummer, grooves well and makes a nice base for all what's on the top but when playing fills the snare sticks out a bit too much.

It's not just a question of mix/sound, real drummer would either rush the fill or play it laid back, the way it is now just feels...hmmm...I don't know how but not as good as it could.

f.i. drummers playing in a big band learn how to prepare for such things, they either play with the band (your case here) or make a sort of an jump-off point so the band bounces off them.

but even in the first scenario the drummer would inject something that'd glue the phrase to the main rhythm. (do I sound conservative here?)

I like what you're trying to pull off but I think it's not there yet, and to be honest, I couldn't program it any better. If you've got time I can ask my drummer to play it to illustrate my point, but that can't be before Saturday night.

cheers

k

[pw]

the production diary is an awesome idea, thanks for going to the trouble!

And, I've got to say that, for me, Day 1 is by far the best.

Apart from the over-production, the improvised nature and the freeform feel of the original vocal delivery is totally lost. From reading your writing process, this is your intention I guess. But, from hearing the progression from the original takes to the final master, it is not just a case of the track being over-produced (with some *awesome* sounds,BTW - love the gtr sound on day 4) but over-rationalized - like the vocals are now in a straightjacket when they were once so free...

anyway, a great look into the production/composition process, thanks again!

[Kim Lajoie]

I know (or at least think I know) what you mean, but I guess what bothers me is this -

loop sounds like a real drummer, grooves well and makes a nice base for all what's on the top but when playing fills the snare sticks out a bit too much.

It's not just a question of mix/sound, real drummer would either rush the fill or play it laid back, the way it is now just feels...hmmm...I don't know how but not as good as it could.

f.i. drummers playing in a big band learn how to prepare for such things, they either play with the band (your case here) or make a sort of an jump-off point so the band bounces off them.

but even in the first scenario the drummer would inject something that'd glue the phrase to the main rhythm. (do I sound conservative here?)

I like what you're trying to pull off but I think it's not there yet, and to be honest, I couldn't program it any better.

Ah, I get it. I'm not too good at programming drums. It's something I've been working at, but I've still got a long way to go. Maybe the only way to get "there" is to spend five year learning to play drum kit (come back to me in 2012!).

If you've got time I can ask my drummer to play it to illustrate my point, but that can't be before Saturday night.

You know, I'd really appreciate that. I'd love to hear a live drummer's response to this.

-Kim.

[bduffy]

the production diary is an awesome idea, thanks for going to the trouble!

And, I've got to say that, for me, Day 1 is by far the best.

Apart from the over-production, the improvised nature and the freeform feel of the original vocal delivery is totally lost. From reading your writing process, this is your intention I guess. But, from hearing the progression from the original takes to the final master, it is not just a case of the track being over-produced (with some *awesome* sounds,BTW - love the gtr sound on day 4) but over-rationalized - like the vocals are now in a straightjacket when they were once so free...

anyway, a great look into the production/composition process, thanks again!

What's wrong with straightjacketing singers?

[Kim Lajoie]

And, I've got to say that, for me, Day 1 is by far the best.

Apart from the over-production, the improvised nature and the freeform feel of the original vocal delivery is totally lost. From reading your writing process, this is your intention I guess. But, from hearing the progression from the original takes to the final master, it is not just a case of the track being over-produced (with some *awesome* sounds,BTW - love the gtr sound on day 4) but over-rationalized - like the vocals are now in a straightjacket when they were once so free...

What do you mean by "over-rationalized"? Is there a particular point at which the vocal changes from free to restricted? Is it the editing, or the arrangement and orchestration of the other instruments around it?

-Kim.

[pw]

And, I've got to say that, for me, Day 1 is by far the best.

Apart from the over-production, the improvised nature and the freeform feel of the original vocal delivery is totally lost. From reading your writing process, this is your intention I guess. But, from hearing the progression from the original takes to the final master, it is not just a case of the track being over-produced (with some *awesome* sounds,BTW - love the gtr sound on day 4) but over-rationalized - like the vocals are now in a straightjacket when they were once so free...

What do you mean by "over-rationalized"? Is there a particular point at which the vocal changes from free to restricted? Is it the editing, or the arrangement and orchestration of the other instruments around it?

-Kim.

I think it is the editing and the arrangement of the other instruments around it. Just that the original has room to breath, and the 'quirkyness' of the vocal is a key element in this context - this to me is a good thing, as a great vocal delivery with unique character is not something to be hidden in a mix or in dense production (IMHO anyway).

I see it is your working process - to get the vocal takes without all the composition stuff getting in the way. But, to me, this seems a great opportunity to excercise restraint - a singer who is able to create an atmosphere, and improvise melodic lines without a backing track - that is gold right there.

[soulata]

If you've got time I can ask my drummer to play it to illustrate my point, but that can't be before Saturday night.

You know, I'd really appreciate that. I'd love to hear a live drummer's response to this.

-Kim.

ok, consider it done. on Saturday.

[Kim Lajoie]

I'm starting a new production diary for our next song "Sunlight". Check it out here

-Kim.

[soulata]

I played Undone to our drummer, he couldn't quite get it, I'll have to send him the mp3.

k

[ahjteam]

Questions and comments marked after the quotes

Day 1: 26/2/2007

Download Audio

This is the very start of the process. Maxine scribbles out some lyrics - I'm actually amazed at how fast she comes up with lyrics that don't suck (It usually takes me over an hour per song - she does it in minutes). I pick a key - any key, whatever I'm feeling like on the moment. Horse Head was in Bm. Undone is in Am. I sequence a bunch of piano notes on A and loop them for about two minutes. Then I get Maxine in front of the mic, and let her wail over that A.

We did a number of takes of the whole song, each with a different singing style. We had "Rock Chick", "Spoken", "Buttery", "Screemo". "Rock Chick" was the first take - just a general slightly-agressive rock performance. "Spoken" was mainly spoken, with a little tonality on a few words. "Buttery" was very light and airy - you can hear it in the first half of the third verse in the final version. "Screemo" was even more aggressive.

This gave me a lot of flexibility in riding the emotional impact of different sections of the song. Most of the song is based on the "Rock Chick" takes. The "Spoken" takes were used in the introduction and the "You turned it on" after the 2nd and 3rd choruses. The "Screemo" takes were used in the background of Verse2, the second part of Verse3, and for the final choruses.

The Day1 snapshot is actually missing a whole lot of these takes. I can't remember if we'd recorded them and I'd just put them out of the way, or if we just hadn't done them yet. You can clearly hear the "Rock Chick" takes, with some of the spoken takes in the background of a few places. There might be a few "Screemo" takes in there too.

-Kim.

* What kind of equipment did you use in the recording? Mic, compressor etc?
* And just out of curiosity: where did you use the screemo, I didn't hear it?

Day 2: 28/2/2007

Download Audio

At the end of Day 2, I'd almost finished "optimising" the vocals with Melodyne. This is where I get the main chord sequences figured out, and match the performances to those chords.

It might seem a bit backwards to record the vocals first, and then develop the harmonic basis later. I can only do this because of Melodyne. It means I can get Maxine recording vocals as quickly as possible while she's inspired and ready, without making her wait for me to do some composing first. It also means I don't have to sit there with her and collaboratively write the melody and chords.

She can focus more on the emotion and performance when she's not worrying about singing the right notes or the right melody. As a side-effect, most of the takes have different melodies and timing, which means they don't line up at all. Sometimes I take advantage of the differences (like in the verses), sometimes I force the takes to conform (like in the choruses). I couldn't do this without Melodyne.

Interestingly, when I played the final version to my sister, she complained that I didn't "fix" her voice enough - that it still sounded like she was a "bad singer". I find this interesting because it's a fine line to tread when correcting vocals. Too much correction and it sounds lifeless. I tried to keep as much of the original performance, making it sound like she just did that performance on the right notes. I definitely fell on the side of too much expression (as opposed to not enough). Maybe next time I'll make her sound like Eiffel 65.

-Kim.

* It gets pretty insane at 0:40 with overlapped vocaltracks.
* What do you mean with that side-effect? you do have that pianotrack recorded seperately that is playing in the background, right? if not, why? also you could've recorded the click track to match the timing

Day 3: 4/3/2007

Download Audio

By Day 3 I'd finished optimising the vocals and applied ballpark eq and compression with Voxformer. I've also added the rhythm section for the chorus and verse 2.

Finished optimising and (initial) processing vocals. Added rhythm section (drums and bass) to chorus and 2nd verse. This is where I lay down the groove and start to establish the vibe of the song. As you can hear, the drums have got quite a bit of swing in them. That's a custom groove quantise I made in Cubase that I use from time to time.

The drums are from Sonic Reality's Studio Drum Capsule, separated as individual elements in SampleTank2. That way I can process each element (snare, kick, hats, cymbals, toms) separately. I usually work with electronic music, so I don't feel any great need to treat the drum kit as a single instrument - I'm not going for a believable "band" sound here.

I like to give the drums quite a bit of groove - even on heavier or otherwise "straight" (non-swing) songs - to juxtapose against the straight 16th of the bass and guitars (and other instruments eventually). It tends to give the song a bit more "vibe" and goes a long way to making it not sound like Rammstein.

As I've written elsewhere, I set the levels so that the snare is roughly level with the lead vocal, and the kick and bass are just a bit behind. I start setting levels and eq right from this early stage for two reasons:

1) It informs the composition process. It's so I can add parts that fit together like a jigsaw puzzle. Often I'll choose register and even harmonic complexity for a part based on the the spectral content of the existing parts. It also helps keep size in perspective - I make sure each part is suitably "sized" (the size of the sound) to work in the mix. For example - I could have made those chorus guitars really massive, but it works better in the mix to have them fit under the snare. Getting it right first time saves me wrestling with it later in the mix, and also prevents me from making wrong composition decisions further down the track based on a sound that may have to change radically to work in the song.

2) It makes it much easier to mix at the end if the levels and eq are pretty close to their final position. It really relegates the mix to "making it sound better", instead of "making it work". Attempting to mix around 60 tracks would be a nightmare otherwise.

-Kim.

* 60 tracks in studio is nothing, if you have the time. If you would have to mix 60 tracks in studio, you would most propably also have the time for it.

* My best/worst is 32 tracks (would've used more if the console had more channels and the venue had more mics) of live-reggae-band. They *were* supposed to be only 6 members, but there was 18 of them, and I only had 30 minutes to mix it in the soundcheck after panicing with the mics etc (because of the time it was more of a linecheck than a soundcheck). Can't remember the name but I think the lead singer was Ras Lawi or something like that and half of the band was from Jamaica, others were locals. 1 drummer, 2 percussionists, 3 sax, 1 trumpet, 5 backvocals, 1 guitars, 1 bass, 1 keyboardists, 1 dj/fx-guy and 2 leadvocals. It was really crowded on the stage and that was the first time when I started to appriciate groups and group inserts.

Second worst was this one thing that I switched with this one guy that was supposed to be a disco. It was not. It was 18 (EIGHTEEN) hiphop-artists and four dancegroups with some really bad RnB music playing in the background and no DJ. I think I was working first time alone there, no monitor guy, no light guy or other technical staff and I had to do a soundcheck for all of them (which was really unnessecary, I used the same settings on all bands) and time to do the soundchecks was less than an hour. About the only thing that was usable in real life that I learned in the military: To work under pressure. You are always in a hurry to wait.

Third worst was Taiga about 04/29/2006: 8 etnobands, all different lineups (no normal g-b-d-v things, but accordions, violins and some really weird percussions). I was in the monitors, but the soundchecks started at 12am because the other guy didn't show up so my boss had to come there, the soundchecks ended at 18pm and the last band didn't even get to do theirs... FYI: The guy who didn't show up was the same guy as the one responsible for the incidident in the second worst thing. He got fired after this incident. Alcoholic bastard.

FYI: MIXING THEM SUCKED! BIG TIME! NEVER AGAIN! Think about the horrors of the monitor guy that 18 member band... *shivers* but the music was ok. One of the worst when I was working at the venue was one I wasn't on, but I heard it really sucked. "Suomen Live Aid"... with 27 performers, which I think was more than the customers :/

Day 4: 5/3/2007

Download Audio

This is where it starts getting really fun. I've worked out the structure for the first half of the song. I've added a bunch more guitars and synths, spot effects and automation. You can still hear the backbone rhythm section from Day 3 though - hear how even with the added instruments, the vocal and snare (and kick and bass to a lesser extent) are still front and centre.

I've added the full four-guitar compliment to the chorus, though it's not easy to pick them out - they blend as a texture, rather than a melodic or harmonic part. One of the guitar parts is a straight Mesa-style overdrive. One part is a bitcrusher going into a vintage amp. The two other parts are fuzzed atonal harmonics. All guitars are processed with Amplitube2.

I also have the Verse 2 violins and violas added at this time. I wasn't sure about this, they sound pretty thin on their own, but I wasn't sure if there'd be room in the mix for a full string section. Later I added cellos and basses in parallel octaves, following the main melody (the violins and violas harmonise it, along with the background harmony vocals). The strings are all from Miroslav Philharmonik, of course.

Notice that in this version verse 2 is twice a long as in the final version of the song. I did a similar thing in Horse Head (a double-length second verse), but this song was getting too long. I think it was threatening to push five minutes, whereas I as was targeting 3:30 (the perfect pop song length). I had to cut something, and the second half of verse 2 was the least necessary (and also means we get the the chorus faster!). I couldn't bear to cut anything else, so the song ended up weighing in at 4:30. It's a bit on the long side, but it covers a lot of ground.

-Kim.

* did you record the guitars or did you use a synth to do them?

Day 5: 7/3/2007

Download Audio

The Cubase project file is taking longer and longer to load. 45 plugins and counting. I usually surf KVR while the project loads. My computer boots up faster than this project.

The song structure is complete - I've finished the third verse, final chorus, and the outro (is that short for "outroduction"?). I don't fade out the outro at this stage - I do that at mastering... so I've got these two bars that just loop on for half a minute.

The 2nd half of the third verse was fun. The lead vocal is not distorted at all. What I've done is duplicated it twice, and processed the two new copes with Replicant. The bitcrusher in Replicant is awesome. Unfortunately the random nature of these non-deterministic plugins means that no playback will sound the same. What I did was bounce down each track three times (so, six in total), and then edited the best bits together. Mostly, it just sounds like horrible digital glitching, but occasionally you can hear bits of vocal in there.

I've also added subtle vocal harmonies to the 2nd and 3rd choruses. They're just the lead vocal re-pitched with Melodyne. It makes the vocal sound bigger by adding additional harmonic content (musical harmony, not distortion!). Compare the vocal sound in the chorus with that in Day 4. You'll notice that Day 4's chorus vocal sounds a lot simpler and more "naked'. I've left the first chorus without harmonies to keep it raw in comparison - supported by the lack of orchestral strings (which further brings out the brutal guitars).

-Kim.

* Even my Cubase-template has 40 plugins (mostly eq, compressor, limiter, sidechain and few guitarsuite plugs), maybe its time for an system-upgrade?
* You also could've used the replicant in a fx-send if it doesn't cause any latency, could've saved a little cpu-power with one less channel used?

Day 6: 9/3/2007

Download Audio

The Mix. Track saturation, panning, reverb. With the composition finished, I can focus on the types of processing that don't change the sound, but make it "better".

As I've written elsewhere, I like to add saturation to almost every sound at this stage. No so much that it changes the functionality of a sound (its place and function in the mix or in the song), but just enough so it adds character and complexity. Of course, saturation isn't the only way to do this, some people use modulation processes like phasers or choruses. Some people use impulse-based colouring. Other people use character compression. I happen to have a taste for dirty, hairy sounds. I like the way it piles up and adds a kind of upper-mid "gzgzgzgzgz" that works really well with already-distorted sounds.

I leave panning to this late stage because I prefer to compose in mono, for a number of reasons. One is that (non-audiophile) people tend to listen in mono, so I don't want to include panning as a significant composition device. Also (and probably more importantly) panning adds extra space in the mix, which I would fill up in the composition process. Once this space is filled, it's extremely difficult to get it back without losing sounds. My composition process is very much one of adding more sounds until I can't fit any more, so I tend to come to the mix with something that's a bit more dense than it should be. Leaving panning to this stage gives me a bit of room to add some space back in.

I'm not a fan of "big mono" sounds - single sounds that occupy a lot of stereo space through the use of processes like phasers and flangers. I can usually fit two or three sounds in the place of a single "big mono" sound... and I much prefer the added harmonic and compositional complexity of more parts over the sonic complexity of a single, more spectrally-complex part. When I use a wide stereo pad preset, I typically collapse it to mono before doing anything else.

Likewise, I don't use autopanners because they take up a lot of space. A statically-panned sound only has to work at a single location. A mobile sound has to work at ALL the locations it might be at. This makes it much harder to mix, and tends to make things a lot more spacially-complex than I like. It's easier to hear subtle sounds if they're not moving about spatially. Once you add autopanning in, subtlety easily gets lost into mush (even more so than normally!).

I also don't use any sidechaining. There seems to be lot of excitement around these parts about side chaining - particularly when compressing the bass and kick. Personally I don't find a need for it. Like autopanning, subtle sounds are easier to hear if they're not fluctuating in volume.

As a side note, I don't remember the CPU overloading like this. Might be something to do with running the project off my (external) backup drive. Cubase is exporting straight to mp3, and the CPU can't handle it so the "realtime" export is sounding like a bad timestretch. I'm actually really surprised the resulting file sounds fine.

-Kim.

* you said you like to compose in mono... do you use mono reverbs too?

Day 7: 10/3/2007

Download Audio.
Music Cafe thread here.

The Final Master. If I ever selling this stuff seriously, I'll have to get my work professionally mastered. Mastering is HARD. The more I learn and the better I get, the harder it gets. It's a completely different mindset, a completely different way of listening. Out of all the stages of creating a song like this, from lyrics to recording, from selecting sounds to composition, from mixing to distribution; mastering is the most difficult task of all.

I mastered this song my lining the mixdown against a reference commercial track, and matching the eq and dynamic range by ear. I also reduced the stereo width of the low end because the room reverb on the kick was blurring the image a little too much.

That's it!

-Kim.

* You should really consider getting a spectrum analyzer. not because your master would sound bad, but it could sound better.
* I don't know if it is the bass or the kick, but it "booms" nastily various times when they play at the same time (eg during second chorus), I heard this when I turned up the volume on my amplifier to the normal listening volume that I use. Thats why you should atleast try the sidechaining
* Kilroys master sounded better atleast on my hometheather speakers
* BTW: Url was broken, link fixed in the post

[Kim Lajoie]

Questions and comments marked after the quotes

* What kind of equipment did you use in the recording? Mic, compressor etc?
* And just out of curiosity: where did you use the screemo, I didn't hear it?

The analogue chain was pretty simple. The mic was a Groove Tubes GT50, running straight into a Presonus Firebox. The screemo is in the second half of the third verse, and in the background of the second verse.

* It gets pretty insane at 0:40 with overlapped vocaltracks.
* What do you mean with that side-effect? you do have that pianotrack recorded seperately that is playing in the background, right? if not, why? also you could've recorded the click track to match the timing

The vocal tracks were overlapping at this stage just while I was recording and organising them. There's actually even more vocal takes in the project that were muted or disabled.

I'm not sure what you mean about the piano track. The "side-effect" that I wrote about stemmed from the fact that there was no pre-written melody that Maxine was singing to. She just made it up as the words came. Every time she sang it, she actually sang a different melody.

* 60 tracks in studio is nothing, if you have the time. If you would have to mix 60 tracks in studio, you would most propably also have the time for it.

Maybe true, but it doesn't mean it'd enjoy it.

* did you record the guitars or did you use a synth to do them?

There are four guitar parts that I played on a real-life electric guitar. There are also a couple of distorted synths in the background to add texture.

* Even my Cubase-template has 40 plugins (mostly eq, compressor, limiter, sidechain and few guitarsuite plugs), maybe its time for an system-upgrade?
* You also could've used the replicant in a fx-send if it doesn't cause any latency, could've saved a little cpu-power with one less channel used?

Well, these weren't all eqs and compressors. There were four Amplitube2s in there, and two SampleTank2s (one fully-stocked). I'm probably due for a system upgrade in about a year. I'm currently running a first-generation dual-core CPU (AMD 4800+, before the Intel Core Duos came out).

Are you familiar with Replicant? Using it as a send would result in a very different sound. I didn't insert it on two different tracks, I inserted it on two copies of the same track. The two Replicants had slightly different settings (one had the low pass filter motion enabled). Besides, Replicant hardly uses any CPU anyway.

* you said you like to compose in mono... do you use mono reverbs too?

No, I mostly use stereo reverbs. When it's part of the final mix, that's when I'm spreading the mix back out to stereo anyway.

On the other hand, "effect" reverbs (like reverse or gated reverbs) I usually do mono.

* You should really consider getting a spectrum analyzer. not because your master would sound bad, but it could sound better.
* I don't know if it is the bass or the kick, but it "booms" nastily various times when they play at the same time (eg during second chorus), I heard this when I turned up the volume on my amplifier to the normal listening volume that I use. Thats why you should atleast try the sidechaining
* Kilroys master sounded better atleast on my hometheather speakers
* BTW: Url was broken, link fixed in the post

I might need to upgrade my monitoring setup. I have a real aversion to "mixing with my eyes".

-Kim.

[stash98]

How long have you been producing? Just wondered because you said you just picked a key. I spend a lot of time finding the singer's key before I go forward. That's more important then anything really. It sounds like you had to do some tuning. You should never just guess a key if you want a great vocal..you need to know exactly which key fits your singer....then proceed.