Composing improvisationally with Ableton Live

I just completed a batch of new music, which was improvised freely in the studio and then later shaped into structured tracks.

https://soundcloud.com/ethanhein/sets/tabla-breakbeat-science

I thought it would be helpful to document the process behind this music, for a couple of reasons. First of all, I expect to be teaching this kind of production a lot more in the future. Second, knowing how the tracks were made might be helpful to you in enjoying them. Third, composing the music during or after recording rather than before has become the dominant pop production method, and I want to help my fellow highbrow musicians to get hip to it.

Some background: Back in the dawn of history, I had a jazz group. We recorded an album in the conventional way: we played a bunch of tunes mostly live, and did a little editing and shaping of the performances afterwards. As soon as it was finished, I wanted to start work on a second one. Inspired by the model of Miles Davis’ In A Silent Way, I wanted to record a bunch of full-band semi-free improvisation, and then use Pro Tools to edit it into new compositions. The other musicians weren’t into it, however, and the band broke up before I could convince anyone. Still, I didn’t lose interest in the idea. I tried it with a few different later groups, with varying degrees of success. Since I got my hands on Ableton Live, I’ve stepped up my producing game considerably, and now I think I’ve finally got this improvisation-to-composition process nailed down. Tabla Breakbeat Science represents the fullest fruition of it so far.

There were four stages in creating this music:

1. I put together some rhythm tracks, simple loops with minimal or no harmonic content.
2. A group of musicians, including me, improvised freely over them in the studio.
3. I mined the resulting recordings for samples.
4. I arranged and processed the samples to make complete tracks.

The sample-mining was by far the most labor-intensive part of the process.

Building beats

Some of the beats were constructed specifically for this project. Others were existing older tracks of mine; I removed the melodies and other foreground elements. Each track is built around a sampled breakbeat. I use some hip-hop and dance standards like “Scorpio” by Dennis Coffey and “Cold Sweat” by Mongo Santamaria. Other beats are a bit further off the beaten track, like “Who Is It” by Björk. A few of the samples I left bare, but most I enhanced in various ways: by adding my own drum machine sounds and percussion samples, by chopping and rearranging the samples, and by applying beat repeat. In the most extreme case, I ran some Aretha Franklin samples through a vocoder and completely replicated all of the drum parts with a sampled 808.

Since some rhythm tracks have basslines and other harmonic content, they determined the key and mode we’d be using to improvise. The tracks that are just bare beats don’t imply any particular harmony, so I was free to just pick keys at will in the studio.

Recording the improvisation

The whole project came about because NYU’s music tech grad students and faculty needed a bunch of multitrack stems to do music informatics research on. I volunteered to contribute because it meant a free day-long session in NYU’s fabulous Dolan studio, assisted by student engineers. Normally I have to engineer my own sessions. It’s quite a luxury to be able to just show up and play, without having to set up mics and run the board.

The band consisted of fellow music tech students whose playing I like: Shashank Aswath, a classical tabla player; Chris Jacoby, a folk mandolinist; and Rachel Wardell, a button accordionist. My friend Roger Bender was there too, and even though he’s a “non-musician,” I had him play atmospheric synth pads on my phone using the amazing Animoog app. I mostly played baritone guitar, along with some guitar and mandolin.

The specific personnel for each track was determined by who was available at what time of the day. We did no rehearsing whatsoever, and for the most part only recorded single takes. The backing tracks were all around ten minutes long, which gave plenty of room to gather ideas. I gave very general direction about how I wanted things to be repetitive and groove-oriented. The other musicians didn’t follow my direction too closely, since none of them have a jazz-funk background. That was perfectly fine; I knew I’d be able to create repetitive grooves later using Ableton.

I had intended to have Rebecca Feynberg sing at the session, but she had laryngitis. A few weeks later, she had recovered, and we recorded her improvised vocals on three tracks. I had Rachel sing on them too, in spite of her protestations that she isn’t a singer.

When he’s improvising on the tabla, Shashank uses Bol, a kind of formal scat singing, to come up with ideas. He considers Bol to be more like throat-clearing than actual music, and he never intends it to appear on the finished product. But I love how it sounds, so I convinced him to let me use samples of it along with his tabla.

Finding the loops

The hardest and most musically creative part of the process was to go through the jams one instrument at a time to identify the samples. I’ve spent the past few years doing a lot of remixing and sampling, and I’ve developed a pretty good ear for a promising loop. I find it easier to dispassionately evaluate random pop songs than my friends’ improvisation. Still, once I settle into it, I find myself in a nicely clinical mindset — it becomes just a bunch of music, some good, some not.

On the first pass, I went through and mercilessly deleted everything that displeased me for any reason: weak ideas, technical flubs, poor stylistic fit, or most often, hesitant execution. Sometimes I kept an entire performance intact, but usually I carved away eighty or ninety percent of each take. Then I went through the remaining music slowly, finding the really strong ideas or passages and creating loops out of them. Most of my loops were of a binary length: one, two, four, eight or sixteen bars. I did a lot of quantizing of the performances to tighten up the groove, and sometimes made major rhythmic changes to the loops by moving the warp markers around.

I have been accused of being a “music fascist” by more than one of my friends. It’s a fair accusation. I know what I like, and I can be controlling about it. But being too controlling can grind the life out of other musicians. I like the edited improvisation approach because in the room with the other players, I can be maximally laid back and accepting, and I can welcome whatever it is that people come up with. I save the merciless judgment for later, when I’m alone with the computer, and I don’t solicit much input for that part.

Building the track

With all of the loops in place, the “composition” part of the process went fast. Before even listening, I striped all of the loops of each instrument so they filled eight or sixteen bar “sections.” I didn’t worry about which combinations of loops were playing at any given time; I just left them in the order they were originally played in. I figured that since everything was in the same key and aligned rhythmically, any combination would sound fine. Then I just listened, listened, listened.

I ended up doing very little moving around of the loops from their arbitrary positions. If something was unsatisfying, I removed or edited it. If something was really great-sounding, I extended its length. I made some effort to keep the energy ebbing and flowing by having the density build and diminish, but I tried not to overthink. Ableton has great effects plugins, and I used at least one on every single instrument in every track. The acoustic instruments all got some multiband compression so they wouldn’t be obliterated by the breakbeats and electronic sounds. I don’t like reverb in electronic music, it sounds corny and dated to me, so for spatialization I used delay instead. I tempo-synced it either to quarter notes or dotted eighth notes for automatic hemiola.

I did some extreme time-stretching in places, slowing the samples down to create textural sounds. Ableton has several different time-stretching algorithms, all of which yield different results when pushed past their intended limits. Using “beats” mode with a short decay time creates a rhythmic stutter. Using “tones” or “texture” mode creates nice granular effects that sound especially cool on vocals. And “complex” mode sometimes produces a lovely digital shimmer — “Who Is It” features a particularly wonderful slowed-down pick scrape on the mandolin.

Philosophical questions

Who composed this music? The simplest answer would be me, since I made all the major musical decisions. A more complete answer would be to include all of the other musicians too, since they supplied the improvised raw materials. But it’s more complicated than that. There are places where I’ve edited and processed the performances so heavily that the end result bears no meaningful resemblance to the original. Should the musician I sampled be considered a co-composer in that case? I don’t have a clear answer.

I also use all those samples from commercial recordings, and legally, everyone with a copyright interest in the sample sources has an ownership stake in my tracks too. This is true even if I manipulated the sample beyond all recognition. If we ever do a wide commercial release of this music, I’ll need to either clear all the samples or replace them with non-copyrighted soundalikes — the latter option is the more likely one. The strange thing here is that composer copyrights apply to top-line melodies and lyrics only. So when I sample a drum loop that contains no whiff of the top-line melody, the drummer on the session has no copyright claim at all on my track (unless the drummer happens to also be the songwriter). In newer pop music, they’ve started recognizing the importance of the beat to the overall song, and beatmakers usually get songwriting credits. But back when all those classic funk and soul records were made, no such consideration extended to the rhythm section.

I wrote above about how much Ableton Live itself contributed to the final outcome. I rely on the presets, since I have neither the time nor the brain space to design sounds and effects from scratch. In particular, the music is heavily shaped by the application of beat repeat, which I use on at least one instrument per track, sometimes several. I set the parameters, but the plugin actually “decides” when to repeat things and how many times. Should the software be considered a co-composer? The person who wrote the software? It’s a puzzler. I’d be glad to hear your thoughts in the comments.