Hearing Voices

May 1, 2008 12:00 PM, By Benjamin McFarlane

DISCOVER THE HIDDEN MUSIC IN SPEECH AND UNLEASH THOSE MELODIES TO TAP INTO YOUR CREATIVE POTENTIAL

Writer's block/creative block is about as common as a cold, and it's no surprise: a 40-hour work week (46 is the average in the United States), straining to meet deadlines and goals, washing the dishes, doing the laundry, maybe even dealing with kids. The left brain gets wrung out, the right brain atrophies and your “real” work suffers. But Remix proposes a fix. There is a tradition — or more exactly, a strain of creative impulse — beginning centuries ago with Czech composer Leoš Janàcek and ending up somewhere between Jimi Hendrix, Steve Miller and the Moog Cookbook that takes its inspiration from direct literal transcriptions of human vocal intonation — and it's fun, too.

Janácek used human speech to capture pure emotion, transcribing the intonations of the Czech language directly into his operas. Thus, a vocalized emotion was frozen in time, re-emerging later note-for-note in the opera performances. Ever since, the idea of melodies from speech has been discussed in scholarly writing on musical composition. Contemporary examples of melodies from speech are more simplistic and often lean toward the goofy or humorous, like Steve Miller's Les Paul catcall in “The Joker” — which, great as the song is, sounds about as pleasing to my ear as sawing a chalkboard in half. Slightly more subtle, and closer to the goal of this article, were Jimi Hendrix's notorious chops on tracks like “Still Raining, Still Dreaming.” Using his wah as a rough but convincing formant filter, Hendrix made his guitar talk, and kaleidoscope-eyed fans listened.

You can easily produce some similar inspiration in your music without any guitar heroism by following these five steps.

STEP 1: LISTEN — GATHER RAW MATERIAL

To start the experiment I needed resources, so I began recording conversations. During shifts at the music store where I work, I left a Zoom H4 Handy Recorder on the sales desk, hoping to capture animated, natural dialog. There are tons of high-quality handheld recorders available for recording sound for music production from Sony, Yamaha, M-Audio, Edirol, Olympus and others. I let the H4 record for a while and burned that file to a CD to listen back. The scene was familiar: A guitar teacher explains a Don Ross lick to the clerks. The shipper and receiver razzes me with an Edward G. Robinson gangster voice. I complain bitterly about the preponderance of overdue rental equipment. Listening back showed me there was genuine potential for following in old Janácek's footsteps. But why restrict the raw material to store chit-chat, or for that matter, to English?

Later I recorded myself with a Sony PCM-D50 reading aloud from The Street by Mordecai Richler and then from L'Étranger by Albert Camus, but my narration sounded embarrassingly contrived. I also recorded a jam session that contained lots of playful banter and musical speech, but to cast my net wider still, I visited the Freesound Project (http://freesound.iua.upf.edu), an online community that publishes sound samples under a Creative Commons license.

A casual search on the Freesound Project for vocal field recordings took me on an absorbing trip through a colorful multicultural, multilingual urban landscape from all over the world. In Mexico, a young boy calls out to a friend twice, the second time sharpening his phrase a minor third; a French lady laughs as she tries on clothing, “Ça fait classe, non? Ça fait chic?”; a young woman greets a friend on her cell phone during the Chinese New Year; and Swedish children react to the pealing of the Civil Defense Siren test. Now that was more like it, but I still had my work cut out for me.

STEP 2: UNDERSTAND — DECIPHER PITCHES

Transcribing spoken phrases into melodic phrases is frustrating for a few reasons. First, the speed of delivery makes the fundamental pitch of most syllables impossible to hear. Second, the habit of perceiving speech as tonally neutral complicates things, too. When listening to speech, we interpret meaning and emotion; we don't pause to savor how the spoken phrases fit together melodically. If we did, then maybe we'd see soloists reciting the Gettysburg Address on violin. Third, the syllables in a spoken phrase slur together more than if the phrase is sung to a conventional melody. Thus, hearing the notes clearly enough to find the right melody is a difficult task.

To get around the difficulties, I enlisted the help of the Celemony Melodyne plug-in, which plots the pitch of samples onto a piano roll and provides a score (see Fig. 1). If you sight read, Melodyne's score helps, but I noticed that the melody I perceived didn't match up with Melodyne's score. That's because instead of organizing each syllable into separate notes, Melodyne groups proximal syllables together as one note and plots the pitch as “drift,” like it would if the vocal performance contained an operatic vibrato. Hence, the best visual aid was the pitch-modulation curve with the piano roll.

The first thing to do in Melodyne after the audio is loaded is to select Melodic from the Algorithm menu to make sure that the pitches are detected. If the algorithm is set on the other choice, Percussive, then no pitch data is plotted.

In the Fig. 1 screenshot, my sax player friend says, “Yeah, we play fairly regularly.” Melodyne sees “yeah, we,” “play,” “fair,” “ly” and “regularly” as separate notes. I get a rough idea of the pitches by following the curve.

If you do not have Melodyne ($299; www.celemony.com), a similar process is possible with a score editor such as Sibelius 5 ($599; www.sibelius.com), Make Music Finale 2008 ($600; www.finalemusic.com) or plain' ol' Finale PrintMusic ($99.95). Melodyne is probably the simplest tool because it plots pitch, and in the case of using a score editor, you must set a tempo that corresponds to the spoken phrase. Failing either of those options, your best bet is to slow the phrase down using a program that stretches audio and use your ear. (Note: Near press time, Celemony announced Melodyne 2, which will be available this autumn for $399. Its major innovation is a feature called Direct Note Access, which, for the first time, allows you to edit individual notes within polyphonic audio. Remix will cover this earthshaking breakthrough in detail in upcoming issues.)

STEP 3: REPEAT WHAT YOU HEAR

Listening back to voice recordings wasn't very exciting, but the moment of revelation came when I tried to play them as melodies. At first, I played them on the guitar, and I smiled as I began to move across the fretboard with movements that were completely unfamiliar.

Fun as that was, I wanted multiple timbres to match multiple voices, so I used a MIDI controller to trigger VST instruments. I tried a few of my regular synths before finding Helix, a free soft synth for Mac or PC by San Francisco company Audjoo (www.audjoo.com). Up until now, when someone asked about a good free VST synth, I always said Green Oak's cross-platform Crystal (www.greenoak.com) was one of the better ones. Now I've got a fresh answer. Helix is fantastic. Audjoo's Website is meek and unassuming, but what an instrument! The oscillator wave-shaping and morphing functions, and the overdrive sections quickly became my favorite parameters to tweak. In a tab behind the oscillator section is a vast routing matrix and sequencer for modulation effects, but for this project I just wanted bare timbres. Even if it weren't free, Helix is an outstanding soft synth, and I'm damn happy to have found it.

Using Melodyne as a rough guide, I reconstructed the rhythm and pitch of “Yeah, we play fairly regularly” via MIDI using the Helix synth. Re-creating this sample, and the many others I recorded, required a lot of time and diligence on the keyboard. It was like trying to pick out snippets of a Miles Davis solo. The movements were difficult, counterintuitive and often involved fast triplets and runs that used bizarre intervals and key changes.

I am proficient on guitar, but my keyboard technique is only as advanced as it needs to be to program MIDI. My humble chops were getting a formidable workout, which I had not anticipated. I could not play most of what I was hearing, and in effect, I was learning a set of custom keyboard rudiments. For hours, I deciphered and practiced phrases, and I was learning new tricks. Though I knew my playing was probably improving for the effort, it was drudgery, and I didn't know how or whether this practice would pay off.

At least when learning from someone else's chops, you know what the fruits of your labor are likely to be. You know what phrase you are learning and you can put it in context. Practicing these phrases was like walking blindfolded, and it took some time before I figured out just how each riff could be useful.

After a few days of practicing, I had a small repertoire of odd musical phrases derived from speech, such as, “Yep, we'll see ya soon,” “El sonido debe estar libre” and “Eh, Benny, kid, seeee, eh?” Some of them were useless dead ends. But not surprisingly, several of the melodies evoked complicated musical ideas.

Many of the simplest phrases were best. “Yeah, we play fairly regularly” would sound nice with a bluesy funk feel to it, but originally it contained a vague triplet shuffle. Other phrases were equally suggestive but in completely different faculties. The sentence, “I don't think it'll be too, too bad” had a great descending melody, with some 16th-note triplets and an unresolved finish. It sounded good, and I learned to play it; however, incorporating it into a song as a hook was impossible. Instead, it would make a fantastic breakdown in a transition from G# to C. The phrase, “Are you speaking figuratively or literally” evoked a sinister transition from C# to G — the devil's interval — and behind it I imagined a fat Jamaican bass line. The sense of drudgery was beginning to lift and little light bulbs were blinking on.

Though I had hoped for melodies with enough common ground to produce a single piece, what I had was a collection of very distinct phrases that did not relate to each other in rhythm, feel or key signature. But how could I complain about getting many ideas instead of just one?

STEP 4: ADAPT PHRASES TO A MUSICALLY INTELLIGIBLE RHYTHM

This is the easy part. The humble beginning of “Yeah, we play fairly regularly” turns out to be a pretty hip sequence of notes. Taking Bb as the root note, I zeroed in on the words “fairly regularly,” tacking on a C#. When I got the hang of the movements on the keyboard, I set the sequencer to about 85 bpm and tried recording the phrase a few times. When I had a few takes that I liked, I tried adding chords, and the riff expanded.

STEP 5: FINISH THE JOB — COMPOSE YOUR ACCOMPANIMENTS

By this point, your muse should be well-primed. I finished off the “fairly regularly” track by adding a bass line and drums in Ableton Live. For fun, I brought in several raw recordings from the Freesound forum and from my own portable recorder. Leoš Janøcek was no fool; as I expected, fine music can be derived from simple everyday spoken melodies. What I hadn't anticipated was a musical workout. I think the greatest reward I took from this project was learning new moves.

To learn complex phrasing on any instrument, a best practice is simply to copy the masters. However, by copying speech, from what I experienced in this project it's possible to build a very unique melodic vocabulary from the ground up. If I wanted to become a virtuoso on any instrument, practicing melodies from speech would be a large part of my regimen. For now, though, I gotta keep payin' those bills — back to the rat race.

Listen to an example of a speech sample, its melodic translation and the finished track at Remixmag.com.

Want to use this article? Click here for options!

I want to SUBSCRIBE to remix

REMIX RESOURCES

Download PDF files of glossaries, charts and mixing tutorials to hang up in your studio as quick-and-easy references for your recording process.

POLL QUESTION