FIG. 1: Butt splices align two segments of audio end to end. The butt splice on the left occurs at a zero-crossing, so no pop or click should occur as it will with the edit on the right. FIG. 2: Equal-power crossfades frequently work best for seamless musical transitions. FIG. 3: Like other audio editors, Steinberg Wavelab offers pitch correction. You can shift a note up or down as many as 36 semitones in intervals of one cent (one hundredth of a semitone). FIG. 4: Retiming a drum track can be time-consuming, but it is the best way to tighten up a shaky performance. Tools such as Pro Tools' Beat Detective (shown here) make the job easier. FIG. 5: Playlist editing can bring fresh perspectives by allowing to quickly rearrange the order of musical sections. In Sony Sound Forge, the Playlist can be converted to a new file. FIG. 6: Mixing segments of different sounds is a great method for designing unique effects. FIG. 7: Grafting a new attack onto an existing sound can add punch to your audio. FIG. A: Engineers have been editing audio since the early days of music production. When an edit point is t the beginning or end of a song, white tape (known as leader tape) is used.
Editing has been a part of music production for 50 years, but the advent of ubiquitous nonlinear hard-disk editing has changed the face of music, for better and for worse. Systems like Nuendo, Pro Tools, SAW, and every other DAW give us unprecedented speed, accuracy, and flexibility in editing audio. We can fix mistakes, adjust timing, clean up recording flaws, experiment with rearranging the sections of our songs, and generally fly things around and chop them up in unexpected ways.
Enhanced efficiency is perhaps digital audio's greatest boon. One of the most important efficiencies, a feature used so commonly that we take it for granted, is the ability to make sophisticated edits quickly and to a level of precision that analog just can't touch. Even among some die-hard analogheads, it is now common practice to lay parts down to 2-inch, 16- or 24-track tape, then fly the tracks into a digital audio workstation for editing.
But we can take editing to obsessive extremes, robbing tracks of their feel as we try to attain a mythical perfection (see the sidebar "Overdoing Perfection"). Music producers routinely use editing as a crutch for subpar musicianship, fixing shaky timing and replacing wrong notes, cobbling together solos and vocal performances from dozens of takes, and disguising a band's sloppiness by painstakingly lining up and retiming the various tracks after recording.
In this article, I'll examine a host of editing techniques that you can use to change and improve your music production. I will also focus on some tricks of the trade to increase your editing efficiency. You can take a quick look at how the track-editing process developed in the sidebar "A Brief History of Editing."
The simplest form of edit involves placing two regions next to each other in time so that one region abruptly stops as the other begins. This type of edit, called a butt splice, is frequently all you need to get the job done. Butt splices have two disadvantages, though. If the splice does not take place at zero-crossings of the waveforms, a tick or pop is often heard (see Fig. 1 and Web Clips 1and 2). In addition, this type of immediate, direct transition can often sound jarring or rough. This is where fades enter the picture.
There are three types of fades: fade-ins, in which the signal grows in amplitude from silence to its nominal level over a period of time; fade-outs, in which the signal gradually drops to silence; and crossfades, which are a combination of the first two. Crossfades smooth over edit points by fading out the old material and fading in the new simultaneously. Crossfades can last from a few milliseconds to several seconds and require sufficient material on either side of the edit point to work with. Crossfades can be used to segue smoothly from one song to the next, to fix mistakes in musical performances by allowing notes to be replaced seamlessly, and to create strange new sounds by grafting different elements into a unified whole.
Fades and crossfades have amplitude curves, which specify the rate of attenuation over the length of the fade. The simplest example is a linear curve (which is actually a straight line), in which the rate of attenuation stays constant over the length of the fade. Linear fades are perfectly useful, but they are far from being the only option. Logarithmic fades start quickly, then slowly taper off toward the end. Exponential fades start slowly, then move quickly toward the end. S-type fades start quickly, slow down toward the middle, then speed up again toward the end. Each of these fades has a subtly different effect and is useful in various situations.
Many DAWs allow you to create custom fades by adjusting these basic fade shapes or by drawing your own free-hand fades. When creating crossfades, you can specify the shape of the fade-out curve and fade-in curve separately, although symmetrical shapes are commonly used. Linear (or equal-gain) crossfade curves are often specified as the default crossfade type. In my experience, though, equal-gain crossfades often leave the impression of a hole or dip in amplitude at the edit seam. Logarithmic (or equal-power) crossfade curves don't exhibit that characteristic, because the in and out regions are both being played at higher amplitudes around the edit (see Fig. 2). This is the curve I use as my default curve.
Many songs end by fading out over a vamp or repeating chorus. Pay special attention to these fades, because they play an important role in the listener's final impression of the song. Start by thinking about where you want the fade to end. I tend to like fades that end at a musical boundary, such as the end of a chord progression. Another effective technique is to fade gradually over the length of a chorus, then end the fade after the beginning of the next chorus. This creates the impression that the song is continuing on and on.
When creating song fade-outs, use a mouse or control surface to create fader automation; don't create a computer-generated fade region. The manual fade has a more human, musical character, making the fade feel like part of the performance.
Some vocalists are lucky enough and talented enough to be able to deliver a flawless performance of a song in a single run-through. Those of you in that category can skip this section; for the rest of us, the art of vocal comping (or compositing) comes into play. Comping involves recording several takes of the tune against the backing tracks, then selecting the strongest elements and combining them into a composite performance.
The lead vocal of a song is almost always its primary focal point, so it's very important to assemble the most in-tune, mistake-free performance you can. However, musical perfection must always be balanced against spontaneity and emotional delivery, both of which tend to get lost after too many takes. Having too much material to choose from can also slow the creative process to a crawl. So when you're preparing for a vocal-comping session, consider limiting yourself to a maximum of four source tracks. If the vocalist delivers more than four takes, choose the best four to work with and throw the rest away.
Next, line the tracks up in your DAW. Listen through the takes a verse at a time. Find the one you like the best and use it as your primary track, replacing elements with materials from other takes as necessary. The amount of detail you apply to the vocal edit is largely a matter of personal taste. Musical genre, skill level of the vocalist, and the producer's working style all factor into your decision-making process here. Some people like to choose whole sections of music, some work on the level of lines and phrases, some get to the level of words, and some people even edit syllables of words from different takes. How you work is up to you, but remember that pitch perfection can come at the expense of musical flow.
When the red light comes on, even the most advanced musicians make mistakes. This is especially true in improvised parts and solos. I feel that the most important thing to capture is the musician's energy in a spirited performance, so I don't let the occasional flub worry me too much. However, there are times when a beautiful solo is marred by a particularly egregious honker. If a note is a bit sharp or flat (within a semitone or so of the desired pitch), you can use pitch shifting to correct it (see Fig. 3). The problem is that significant pitch shifting disrupts the delicate attack transients that are so important to a sound's definition and also changes the formant relationship of the various overtones that define the instrument's timbre. Therefore, it's a good idea to try replacing a bum note with another note from elsewhere in the track. This treasure hunt does not always pay off — you need to find a replacement note with appropriate pitch, duration, and amplitude. Longer notes can be used to replace bad notes if you trim their duration carefully.
When you find a good substitute note, paste it into an empty track directly above the bad note. Zoom in to the beginning of the waveform and line up the attack portions of the notes as closely as possible. Drop the replacement note onto the old note's track and create crossfades before the attacks of the new note and the note that follows.
Drums are one of the most difficult instruments to record well. A series of distinct sound sources emanate from different locations in proximity; multiple mics and tracks are often used to allow greater mixing control. Every mic picks up sound from all the sound sources in the kit — the snare drum bleeds into the kick-drum mic, for example. Thus, any EQ or mix decisions applied to the kick will be reflected to a lesser degree on the snare drum on that same track. The snare information in the kick-drum mic will also be slightly out of phase with the dedicated snare track, smearing the all-important attack transient. Multiply this effect by 7 to 12 mics, and a multitrack drum kit can quickly become a dull, thuddy, boomy mess.
This problem can be ameliorated somewhat by carefully editing out material on tracks for drums that are not being struck at a given moment. In practice, this usually works best with tom-toms, since they are struck less frequently than the kick and snare. Listen through the drum tracks for the various tom fills. Delete everything on the track except for the fills, fading the regions you are keeping after the tom fills are completed. Although this may sound unnatural when you listen to the tom tracks by themselves, the stereo-overhead tracks glue the whole sound together, creating a smooth, clean-sounding drum track in which the toms jump right out of the speakers. Gates are frequently used for this purpose as well, but editing by hand avoids false triggers.
The multitrack nature of drum recording makes other types of edits a challenge as well. When moving drums around, you must remember to keep all the drum tracks grouped and edit them simultaneously. A drum edit's biggest giveaway is the unnatural decay of the cymbals at the edit point. You can sometimes disguise this by extending the crossfade region for the overhead tracks a second or so past the edit point. You can also fly in a new cymbal crash right at the edit weld if that makes musical sense.
Digital-editing technology gives users the ability to retime acoustic-drum performances, moving the individual drum hits around in time. You can use this technology to tighten up an unsteady performance or even to change the tempo, within reason. In the past, music producers have used traditional time-compression or time-expansion algorithms to change a drum part's tempo, but this can't correct for a sloppy groove. And this type of digital manipulation can have deleterious smearing effects on nice, crisp drum recordings.
FADES, FADES, FADES
An alternative approach is to chop the drum tracks into a series of small regions, using functions such as Pro Tools' Beat Detective. Beat Detective works by analyzing a drum performance's transients in relation to the tempo of the track, breaking it down into user-specifiable subdivisions of bars and beats. Once the settings are dialed in, Beat Detective separates the performance into dozens or hundreds of individual regions (see Fig. 4). These regions can then be quantized against Pro Tools' metrical grid just like a MIDI drum track would; even groove-template and swing parameters can be specified. Once everything is lined up, Beat Detective applies a series of short crossfades to smooth over the edit points.
If your workstation of choice doesn't offer a feature like Beat Detective, you can chop your track by hand, although the procedure is a little more tedious. Start by using a strip-silence command to break the drum tracks into component regions. You will have to tweak the threshold and region-length settings on a per-track basis for optimal success. Be sure to leave enough time before the threshold trigger so the very beginning of an attack isn't chopped off. You can then use a quantize-region feature to snap those parts in place and a global-crossfade feature to weld the regions together again. Try a crossfade time of 10 milliseconds to start with. Listen through the tracks carefully, and you'll find that you'll have to fix a few things by hand (I end up having to do this when using Beat Detective as well).
To be honest, I've had mixed results retiming drum tracks in this fashion. Sometimes it works beautifully, and other times it makes the parts sound weird and robotic; it varies with the individual performance and musical genre. If the performance feels good but is drifting against the click, there is an alternative to consider. Try breaking the performance into one-, two-, or four-bar pieces. Quantize these larger pieces against the grid. This works because it lines up the downbeat with the click, but the internal structure of the groove is better preserved, creating a more natural-sounding edited track. Resyncing the drums every bar or two maintains the track's rhythmic integrity while retaining everything we like about human feel.
A half-century ago, iconoclastic piano virtuoso Glenn Gould created controversy throughout the classical-music world by heavily editing his recorded piano performances. His finished recordings were composited from multiple takes, which sparked debate over the nature of the recording process itself. Now, structural editing of 2-track mixdowns is common across all musical genres.
Finished tunes are routinely edited for length; entire sections are removed to make the songs more radio friendly. Sections can also be moved around or removed after the fact to tighten a song up or improve its flow. Artists are often too close to a song during primary production to look at large-scale structural issues. Once the mix is completed and some time has passed, the song can be looked at with a new perspective, at which point editing can come into play to improve pacing or momentum.
There are a number of 2-track audio-editing packages such as Bias Peak and Sony Sound Forge that offer a nondestructive playlist-editing feature. This allows you to define musical regions in your mixdown that can then be rearranged quickly within the playlist window (see Fig. 5). Some playlist systems provide crossfades to help smooth out the edit points. If the playlist is not sufficiently sophisticated for a project's needs, you can still rough in your arrangement ideas in a playlist, then implement them in greater detail within your DAW software.
I love recording live performances; there is often an energy, an improvisational spirit, and a lack of inhibition that can't be duplicated when the red light goes on in the studio. But live performances can also ramble: improvisatory expositions that are captivating onstage can feel long-winded when played back in the living room. Aggressive editing can be used to turn 20-minute jams into 8-minute gems.
Start by listening to the raw performance several times. Draw out a rough road map, identifying energy levels, spontaneous arrangements, and soloing instruments. Identify natural areas of tension, build, and resolution. Next, look for obvious places to tighten: meandering sections where no one is really playing a lead, areas of solos that feel aimless or repetitive, and sections where the band is not grooving well. Remove those sections, looking to leave logical in and out points to join the remaining material. Try to sculpt the piece to create a sense of forward motion — of building toward and reaching a climax, then building again.
Making musical edits sound convincing can be a laborious process. Sometimes sections will fuse together naturally, but often they won't — it's the luck of the draw. If an edit isn't convincing, be honest with yourself about it and look for alternative entry and exit points. This process can force you to discard your original notions but can lead down interesting new paths. Sometimes the effect can be spellbinding. With a little luck and some skillful editing, group free improvisations can end up sounding like carefully composed works.
Film and television make heavy use of preexisting musical works, fitting them to picture as needed. In fact, music editing is a particular specialty job within the post-production world. Successful music editors have an encyclopedic knowledge of a wide variety of musical genres, as well as large music libraries available at their fingertips. They can identify the mood for a scene and cut appropriate music to picture, often using multiple pieces from a variety of sources within a single scene. This skill is frequently put to the highest test in movie trailers, where the music editor has 30 seconds to knock the audience out of their seats.
There are times when the edit transition between two disparate pieces or sections feels awkward or forced. A music editor can soften the transition by composing bridge material such as a musical flourish or string pad or by adding sound-effect whooshes or hits at the edit. I often smooth over unconvincing orchestral edits by applying a bit of reverb to the tail of the outgoing piece. This creates a subtle bridging wash for a second or two under the incoming piece.
When you're creating sound for picture, editing is your most important tool — they don't call them sound editors for nothing. Existing sound effects are constantly trimmed, sliced, and diced to fit the duration of events onscreen, in films, television, or video games. But in addition to simply trimming sounds for length, editing can be used as a creative tool for sound design. Different sounds are frequently spliced together to create a new hybrid sound, using, say, the attack portion of an M-16, the body of a shotgun, and the tail of a jet fly by (see Fig. 6 and Web Clip 3). Using this mix-and-match approach, you can generate a large variety of gunshots, explosions, and body falls from a limited set of materials.
According to psychoacousticians, a sound's attack is the most critical component in defining its timbre, so that's where I put a great deal of thought when designing sounds. This approach is particularly appropriate for trying to get punchy weapon fire, explosion, or percussive sounds. Layering sounds is a very useful tool, but there is only so much energy you can build into a single attack transient — digital 0 (that is, all bits on) is still 0 no matter what you do. So when I want to make a really cool, impactful sound, I often take the one-two-punch approach, grafting in a low-frequency thud just before the sharper high-frequency report of the weapon (see Fig. 7). Listeners' ears will fuse this together into a single, seamless sound that is big and mean, without my having to squash the dynamic range through overcompression.
This technique can be used to create new musical instruments for your sampler as well. Years ago, the venerable Roland D-50 created a new battery of sounds by offering sampled attack transients and synthesized body waveforms. It was great fun to graft a flute attack onto a string body or a string pluck onto a horn sounds. You can do the same thing now: try replacing the first 250 milliseconds of a violin sample with the first 250 milliseconds of a piano sample of the same pitch. Use crossfading to blend, and presto, you've created a pianolin (see Web Clip 4).
In editing, speed is power. You can greatly increase your editing speed by memorizing your DAW's edit-keystroke commands. In Pro Tools, turn on Single-Keystroke Command mode (click on the little A-Z button in the upper left-hand corner of the edit window), and keep your left hand on the middle row of your computer keyboard. A and S trim the top and tail of a region, D and G create fade-ins and -outs, and F creates a crossfade in the selected area. Also try turning on Tab to Transient mode, which moves the selection bar from attack to attack with each stroke of the Tab key. Weaning yourself from the mouse wherever possible promotes maximum speed and efficiency.
You can use pre- and postroll to test your edits in context without having to move the selection bar away from the splice point. Setting a 2- or 3-second preroll allows you to concentrate on fine-tuning the edit without losing your place. When selecting large regions of music to be edited, it is all too easy to lose track of your edit-in point while hunting for your edit-out. Dropping a marker at the edit-in point solves this problem handily. Once you've identified your edit-out point, you can jump right back to the marked in point to quickly and accurately create your edit region.
Many digital audio workstations support a global-fade function, allowing you to create multiple crossfades over a large series of regions. For a starting point, try specifying an equal-power curve with a 10-millisecond crossfade length, then listen through for any fades that don't work and tweak them by hand.
In video-game and multimedia work, engineers are often called upon to edit and batch-process hundreds of files at a time. If you don't have time to check each file by hand, you can safeguard against non-zero-crossing-induced ticks and pops at the tops and tails of the files by specifying a 2- to 5-millisecond fade-in and fade-out within your batch processor. AudioEase's BarbaBatch is an example of batch-processing software that supports this very useful feature.
People have been slicing, dicing, and shredding their music for decades, shaping their work in interesting and useful ways. The art of editing is an acquired skill and a critical tool in any audio engineer's bag of tricks. Try some of these techniques in your next musical project and see how editing can clean up, tighten up, and open up new sonic vistas.
A BRIEF HISTORY OF EDITING
In theory, digital editing allows us to produce letter-perfect performances. All the techniques discussed in this article are used every day to create audio illusions: releases that are metrically perfect and far beyond the musical capabilities of the band or musicians that purportedly recorded them. But endlessly worrying a musical track to try to perfect it can lead to sterility and a lack of perspective. And all the editing in the world won't make a beginning guitar player sound like Steve Vai. The question is no longer whether we have the capabilities of altering reality in this way, but whether we should do so. The reaction against false perfection has led to a resurgence in the standard recording technique of yesteryear: recording whole bands in a room together with minimal overdubs. This approach is finding mainstream success in the records of such artists as Norah Jones.
At the 2003 AES convention in New York, legendary producer Arif Mardin discussed this phenomenon in his keynote address. In his opinion, digital tools were all too frequently overused to manufacture performances. In effect, this amounts to the "dumbing down" of commercial music in an effort to pander to a supposedly unsophisticated public. Mardin felt that the public is far more perceptive than the commercial-music industries give them credit for, and that Norah Jones's success shows that honest, heartfelt performance is what ultimately counts.
Analog audio editing has been around as long as analog tape itself and is still in widespread use today. The process is fairly straightforward but demands precision for good results. To start with, locate the edit points by shuttling the tape to the desired location. Then disengage the playback motor and manually rock the reels back and forth until you can hear the edit point line up exactly with the playback head. At this point, place a mark on the back side of the tape with a grease pencil and then find the other edit point, repeating the process. Next, slide the tape into an aluminum tape guide (known as an editing block) attached to the recorder. There are narrow grooves in the block at 90- and 45-degree angles. Line up the marked edit point with one of those narrow grooves (typically the 45-degree angle groove, because this creates more of a crossfade effect), and carefully slice the tape with a razor blade.
Spool off the tape you don't want until you come to the next edit point; one more slice there, and the edited material is gone. The two ends of tape that remain are then carefully lined up, abutting each other in the editing block, and are connected with a piece of cellophane tape known as splicing tape. If the edit point is the beginning or end of a song, the tape is not spliced to more recording tape, but rather to white tape known as leader tape (see Fig. A). Leader tape is nonmagnetic and thus creates no sound when passed over the playback head. Its contrasting color makes it easy to visually pick out the beginning and ending of the song on the reel.
This laborious process has been common practice for 50 years and has been used creatively in countless situations. When recording the Beatles' "Strawberry Fields Forever," John Lennon liked the first half of one take and the second half of another take; he asked producer George Martin to splice them together. Unfortunately, the takes were recorded at different tempos and a semitone apart in key. Martin accomplished this nearly impossible feat by gradually slowing down the tape as it approached the edit point. That edit happens at exactly one minute into the song.
A significant limitation of analog editing is that it is an all-tracks-at-once proposition; it is impossible using standard splicing techniques to move one track in time relative to another on the same tape. So if a bass note is late relative to the kick drum, the best option is to punch in and rerecord it. Necessity is, however, the mother of invention. The need to fix tracks after the musicians were not available led to a high-wire proposition of last resort: the infamous "window edit." This technique involves slicing a "window" of tape that corresponds to the location of a physical track lengthwise down the reel of tape for the duration of the edit, then slipping the window in time relative to the rest of the multitrack tape, then taping it back into its new position with splicing tape. Performing this task in the digital world is trivial by comparison.
Using technical innovation to invent new genres of art is nothing new. In the '50s, pioneering composers such as Pierre Schaefer, Pierre Henry, Karlheinz Stockhausen, and Edgard Varese sliced up recordings of everyday sounds: door slams, church bells, children playing, and so forth. They then recontextualized those sounds by gluing them back together and presenting them as unified compositions. This early branch of electronic music, known as musique concrete, was the first form of music that could not have existed without audio editing. The impact of musique concrete was utterly profound — echoes of this early work have resonated through virtually all electronic music to the present day.
Composer, keyboardist, sound designer, and engineer
lives and works in the Bay Area. He still likes editing with razor blades on an old Ampex 440-B ¼-inch reel to reel. Find him at