Adventures in home recording, Part 2
Last time, I wrote about Track 4 at a fairly high level. But there’s more to its story now, because I edited it.
Editing is a curious thing, especially in music composition. I have long considered the definition of “done” in composition to be when any additional thing you do makes the work worse. In the world of home demos, this has two vectors: composition and audio engineering.
Songs need to be able to stand on their own. As the writer/composer, you don’t get the luxury of being there to give more context or background to what someone is listening to. (Performers don’t have this luxury either, but they might at least be in the room.) Any small part of a song that gives me pause – me, who knows what and why every little detail is there – is going to sound odd to someone else. This song had two of those things.
That major second (shakes fist)
The vocal melody of the chorus phrases were hard to work out. They end with three syllables, and each syllable adds a new voice; so you can think about this as three voices starting on one note, followed by two voices on one note and one voice on a different note, followed by three voices all on different notes. Architecturally, this is satisfying and in instrumental writing it isn’t uncommon. But in this case, that third chord has two intervals: a minor third and a major second (with the major second being the root and seventh of the chord). In terms of theory this is fine, but in listening to it, it sounds like a mistake.
This was an easy fix: find where that one track switched to that last syllable and trim it out (meaning: cut off the waveform before that sounds, and adjust the volume fade out so it doesn’t sound abrupt). The result is less challenging for the listener, and most people probably won’t notice that a third voice disappears from the unison that precedes it.
The outro transition
it’s possible that there’s one final edit I’d make to it (making the last “in” supported by a full six-count measure before the final “time” is sung)
That should’ve been my first clue that I wasn’t satisfied with how this turned out. The transition from the post-bridge guitar riff (in 7/4) to the unison outro (beginning in 6/4) was... sudden. Admittedly, I had gotten decent guitar takes and I just copied and pasted those, so there was no architectural transition, just smoothing the audio. The vocal melody here is a three-part cascade, where the voices complete the phrase “falters, falls and dies” in pieces, with voice 1 singing “falters,” voice 2 singing “falls and dies,” and voice 3 overlapping voice 2 with its own “falls and dies” before they sing “in” in unison. The timing of cutting and pasting 7/4 measures yielded a half beat for all the voices to sing “in” and that goes by so quickly that the following note (the final syllable of the vocal melody, which is held for two bars) appears suddenly, making the transition feel abrupt. Not seamless.
So my first solution was to end the preceding bridge with a new section, echoing the first bridge in the song, which was a singleton. I mistakenly thought this was simple, even telling someone this week, “the answer was right there, staring me in the face. It was even in the same key and same time signature” (though it was not actually in the same time signature, the first bridge is in 6/4; also, the last “in” needs to be supported by a full eight-count measure, so apparently I can’t count correctly in this sea of time changes). I spent an evening trying to insert an entire section into the song before the outro. I threw this option out because a) it provided no harmonic motion, b) it interfered with the emotional arc, and c) it didn’t address the root problem, which was the transition from 7/4 to 6/4.
In my head, the simple solution was to extend the space for that one syllable “in” by adding a single beat. In practice, this means a series of awkward changes:
- Because I can’t figure out how to change the map’s time signature while preserving the rest of the arrangement alignment, I needed to insert one measure of 1/4 at the point of transition between the measure of 7/4 and the measure of 6/4.
- Because inserting one measure of 1/4 cuts across recorded waveforms, those waveforms are cloned (this would be relevant later).
- Because I did not write down what the exact settings were for guitar and bass and organ everywhere I used them, matching tones will be difficult (or impossible) if I need to re-record anything.
- Because of that, I copied parts of the guitar and bass waveforms to capture their one note or chord that needed to be repeated, copy and pasted them three times each, and stitched them together.
This worked well enough for all of the guitar parts; however, because they were cut and pasted in their original riffs (from which I was cut and pasting them again), there was no variation in the waveforms at all. So a series of sneaky volume automations across the repetitions tries to make them sound not identical. This approach couldn’t work for the organ track, so that needed to be re-performed to match the new timing. Fortunately, I’ve spent a lot (a lot) of time with organ patches and recreating this – which could be a nightmare – was pretty easy for me.
The vocals were the hardest part, because they needed to be “bent.” (In my DAW, “bending” audio waveforms is stretching them internally – making them longer here, shorter there, etc. To do this, you need to fix endpoints, specifying the confines of the part of the waveform you’re going to bend; this ensures you don’t inadvertently change things that come before or after your edit.) But because the time insertion (again, one measure of 1/4) cut through vocal line waveforms, the DAW cloned them. Which meant that when I went to bend the “in” to be longer, extending it in time to the right, I couldn’t remove the excess squished to the right because it would pull the other side leftward in the subsequent waveform.
It gets worse. I also doubled every vocal take in place (meaning: two waveforms on top of each other on the same track). That means every cloned vocal track results in four waveforms: two before the split, and two clones after the split. That means that for this one syllable of “in” to get longer, there were three tracks, each with two waveforms that were cloned (so, four waveforms each) that needed to be adjusted. That’s twelve separate edits that need to be as identical as possible. To be able to do this cleanly I needed to:
- Put both waveforms of the take following the bent part, which would be ostensibly unchanged, on their own (new) track in order to isolate it.
- Bounce each waveform in place to separate them from their clone (so, if “in time” is split between words, both “in” and “time” need to be bounced – and because vocal takes were doubled, there were two waveforms to be bounced for each). And to bounce them, I needed to separate each double to a new track, change it, and put it back with its partner.
- For each waveform that was to be bent, extend them the same way at the same points, squishing the excess to the right, and then trim the waveform to cut off that excess (everything after the bent syllable ends).
- Crossfade the doubled bent syllable (on one track) and the doubled syllable that follows it (on the new track).
Tedious? Yes. Did it work? Also yes. I have to say, the bending feature in my DAW (Presonus Studio One) worked like a charm for this. It does sound a bit like the record skips underneath the vocal, but I’m also not sure that’s a bad thing – and I’m not sure that re-recording the guitar would not have also sounded like that anyway.
Lastly, I modified the drum track to be a bit more organic in the transition measure (grr, one beat), and I panned the “new” vocal tracks I added wider to make them easier to hear (an option I only had because of that edit process, so that’s a happy accident). This solves the flow and transition problem, and now nothing sticks out. We can call this song complete and stop thinking about it. Progress!
Listen to track 4, “black and gold” on soundcloud