Digital Audio Workstations:
Working with ProTools Free
Getting Started | Setups & Troubleshooting | Basic Production | Voice Editing

by Jeff Towne

The most common activity involved in making radio programs is editing voice tracks. Oddly, many of the maneuvers that seem the most natural to us are rarely performed in the music production world, for which ProTools is primarily designed. Sure, music producers do plenty of "comping" of vocals, assembling a final performance from several takes, but there is rarely a call for the kind of deleting, condensing or reordering of voice tracks that we do in radio every day. So there are not many demos about this in the ProTools manual, but luckily, the program allows some sophisticated editing that will quickly banish any fond nostalgia you may harbor for the days of razor blades and spicing tape.

The beautiful thing about ProTools is that there are several ways to perform a function, so you can find a style of working that fits your rhythm best. There are plenty of other routines beyond what we'll cover here, and if there's a technique you think is really great, please let me know about it. Once you have recorded your dialog, whether it's narration that just needs to be cleaned up, or interview actualities that need to be condensed and reordered, the procedures are largely the same.

(Note that the keyboard shortcuts mentioned below are Macintosh keys, for Windows, substitute "Control" for Command and "Alt" for Option.)

Here are the major techniques:

• Separate Region
• Region Delete
• Trim
• Real-time select and delete
• Nudging

But first, there is one technique that should be really great, but with which I can never get any useful results. It's called "Strip Silence" (under the Edit window) and will search your soundfile for areas of silence, which can be custom-defined by audio-level threshold, and other parameters. But, I can never actually get anything useful out of this function. I can get it to delete either too much or not enough, to either make too many segments or none at all. If you are using this feature successfully to rough-out dialog, please drop me an email and let me know how you are doing it. I think it may be better-suited to breaking a drum grove into component hits than defining narration takes.

So we'll just "cut" manually. Depending on what type of editing you are doing, the editing mode (Slip or Shuffle) makes a big difference. The F1 key will drop you into "Shuffle" mode, the F2 key into "Slip" or you can simply click the appropriate box on the upper left of the control bar. Using "Shuffle" when deleting audio is analogous to cutting out a piece of tape and splicing the surrounding ends together. Using "Slip" will simulate cutting a piece of tape and splicing-in leader tape to keep the sound bite separate.

Separate Region

One useful and straightforward way to edit your voice tracks is to play the soundfile until you find the spots you want to make cuts. Zoom in on those spots, then using the region tool, click with the mouse once at the spot where you would "cut the tape" keeping in mind that you will reduce clicks and pops if you place this edit point at a "zero crossing" where the waveform crosses the horizontal axis of the display. Choose "separate region" from the file list (Keyboard shortcut Command E). After you have made new regions at the beginnings and ends of the desired sections, you can delete, or move any of the segments you have created. Barrett Golding has explained this style of editing very comprehensively here.

Region Delete

Instead of zooming in really closely to separate regions, it's often much quicker to do some rough cutting by eye, and then going in and cleaning things up. Even from a quick look, it's pretty easy to see where the rough beginnings and ends of these actualities are within the larger soundfile.

This graphic uses red arrows to indicate the spots where the actualities start. The beauty of non-destructive editing is that, unlike cutting analog tape, after we've thrown audio away, it's very easy to get it back again if we've made a mistake.

In Slip mode (F2), we'll use the region tool (command 3) to select the silence and unwanted audio, and coarsely chop the larger soundfile into individual actualities. Click and drag with the mouse over the areas of unwanted audio (don't worry about being too precise yet) and hit "delete."

After you have individual regions, you may want to use the "grabber" tool (command 4) and double-click on the region to name it something distinctive.

Now, let's zoom in and be more precise. It's important to be able to see what you are doing, so get used to what the various viewing tools will do.

Horizontal Zoom with outer arrow buttons (or command bracket left for out, right for in) Adjust waveform height with upper and lower buttons on the left (or command option bracket)

Use the Zoomer tool (command 1) to zoom-in on the beginning of the region. Either drag with the zoomer tool over the range you want to see, or click at a point on the waveform. The display will zoom in one level every time you click with the mouse, centered on the point where you clicked. Additional clicks will zoom closer. Option click will zoom back out. If you have a region selected, Option F will zoom so that region fills the display. Adjust the waveform height carefully so you can see low-level audio.

Once you are zoomed to a level where you can see the waveform well, use the Region tool (command 3) and drag over the unwanted section, then hit delete. Or use the Trimmer (command 2) and click on the edge of the region and drag back to where you want the region to start, or simply click at the point to which you want to trim. If you have accidentally chopped-off some audio you need, simply use the trimmer tool to drag the edge of the region back out to the proper point on the soundfile.

Then, LISTEN TO IT!!!! And just in case you weren't paying attention:

LISTEN TO IT!!!

This is the biggest mistake people make: editing with the eye rather than the ear. It's easy to accidentally trim off an initial "f" sound, or some small sound that doesn't LOOK important, but plays some crucial role in making the voice feel natural. So listen.

Trim up the end the same way, and, you guessed it: listen.

Real-Time Select

A very natural-feeling technique for editing dialog is to mark a region in real-time. To do this, press the spacebar to play your soundfile, and when a section you would like to delete begins, click the Down Arrow on your keyboard (you won't see anything highlighted yet, that's OK). At the end of the section, press the Up Arrow. You will have defined the region in a rhythmic, auditory way, not purely by sight. Do NOT stop the playback, you will lose your selection. While the cursor is still moving, hit delete, and you've made an easy edit. With a little practice, you can synchronize to the rhythms of speech and will be cleaning up voice tracks with alarming speed.

This technique is also very helpful for trimming the ends of actualities: instead of always zooming in and trying to find the right point by dragging a mouse, let the voice track play, and hit a Down Arrow where it sounds natural to end. Hit an Up Arrow anytime later, and in Slip mode, delete the selected region. You've quickly trimmed the tail.

When making a selection, especially in real time, you might discover that the edges of the selection are not exactly where you want them. Use the left arrow on the keyboard to scroll the display automatically to the beginning of the selection (the right arrow will scroll to the end of the selection.)

Then while holding the shift key, click and drag the edge of the selection with the region tool, until the boundary is where you want it. Alternately, you can drag either of the blue arrows that appear at the top of the timeline.

Wink Wink, Nudge Nudge, say no more...

If the selected region is the correct length, but it a bit too early or late, you can nudge the selection range, while leaving the audio in-place. With a region selected, hold Shift and then press the Plus or Minus keys to nudge the selection backward or forward. The nudge value is controlled by the same box as for nudging audio: in a small drop-down menu in the upper right hand corner of the edit display. Holding the Option key while dragging one of the blue arrows in the upper timeline allows you to drag the selection range.

Hold Option Shift then use the Plus or Minus to nudge only the beginning of the selection range, while Command Shift and Plus or Minus will nudge only the end of the selection range. Both of those operations will change the length of the selection.

Shift +

Similarly, with an entire region selected, the audio can be nudged within the region boundaries by holding the Control key while nudging with the plus or minus keys.

Try each of these a few times. With the right nudge value, these can be very powerful aids.

On the Level

Go through your voice tracks, using the techniques above to trim out the sections you don't want. You may find that the volume of the track is not as high as you'd like, or that the different sections of an actuality are at incompatible levels, making the edits sound "bumpy". There are a couple of things you can do about that.

You can automate volume to "ride the fader" over the edited audio, evening-out the levels. But there could be abrupt volume changes at edits that cannot be adequately compensated for through this technique. This works well for a general smoothing. This is best left until editing is complete, because further edits will affect the automation data as well, requiring some tweaking of that data. For quick tips on using volume automation, see the Basic Production column here at Transom.

Another way to smooth the levels is to make new regions of sections of the audio that are of a similar volume, and "normalize" each of them separately. The Normalize command is in the Audiosuite menu, and it creates a new file, using more disc space.

You can normalize several segments at once, with each processed individually. Select all of the contiguous regions by clicking with the grabber tool on the first or last region, then shift-click on the opposite end segment. Then under the Audiosuite menu choose: Normalize.

Be sure to pick "region by region" and "peak on each track" then click "process." If you left the upper right control of the normalization window highlighted at "use in playlist" the newly normalized regions will drop into your playlist right where they belong. If you don't choose that option, the new normalized regions will show up in the "clip bin" on the right, but you would have to drag them into your project to use them. If you are feeling really lucky, you can choose "create continuous file" instead of "create individual files" which will result in one solid file (like a bounce) which is easier to work with. But it also removes your ability to tweak individual regions or region boundaries if the transitions don't sound right.

Before Normalization

Normalizing will automatically bring the loudest audio peak up to the highest possible level. The volume of the regions shown below are higher and more similar to one another after normalization. Normalizing will not always give good results, the natural patterns of human speech sometimes require that some phrases are louder than others. But this technique will often get you very close, quickly. Compression applied at the mixer can smooth levels too, but it's often more effective to get the soundfile evened-out, so the compressor doesn't have to work too hard, which can sound unnatural.

 
After Normalization

If normalizing isn't sounding right, under the same Audiosuite menu, there is a similar option called Gain which allows you to adjust the levels of a segment up or down by an amount you determine. Use the "find peak" button to see how much gain you can apply before you would clip the signal. If the peak is -3.3 dB, you can apply that much gain before distortion, although it may sound better to gain it up less, or even drop the gain. The Gain command can help save edits that would not otherwise be possible. Often an edit sounds unnatural because the voice volume has changed from the pre-edit to the post-edit point. Try raising or lowering the gain of one side of the edit, often the transition will sound believable after adding or subtracting a dB or two of gain to one side of the edit.

Try not to normalize or make gain changes many times to the same file. It's better to "undo' and try a different setting than to normalize, then reduce gain by 2 dB, then increase by .5 dB etc. Each process will write more soundfiles, and also will build up dither noise, and your soundfiles may sound noisy or dull with repeated processing.

The file to the right is a "bounce' of the segments above it, creating a new, more even-volume soundfile. You don't have to bounce to a new file (and as with gain changes you shouldn't bounce many times) but it is sometimes easier to manage and move one file around in a mix rather than a string of little segments.

After Bounce

Click and Clack the Edit Brothers

The one glaring hole in this discussion is the problem of discontinuous waveforms. As we have been happily deleting regions and changing levels, we may have created clicks, pops, or thumps at the edit points. I prefer to try the edit first, often one gets a satisfactory result without laboriously searching-out "perfect" edit points. But if you do end up with clicks at edits or at region boundaries, there are a few ways to deal with them.

This edit (the vertical blue line) created a click. You can see that there is a discontinuity in the waveform from the peak before the edit, to the lower level after.

In Shuffle mode, use the trimmer to move the edit points on both sides to a "zero crossing" where the waveform crosses the horizontal axis. This is now a smooth edit.

Sometimes even such a small adjustment of the edit point ruins the edit, so in that case, try a very short crossfade. Select a small region across the edit point.

Under the Edit menu, choose Fades>>Create Fades (or type Command F)

This will smooth the transition, and should remove the click. If this fade does not work, select the fade and delete it, and try again with a longer or shorter fade. There are many options in the fade window, but for such a short transition, the default fade should work as well as any. (More about complex fades in a future column...)

These short crossfades can smooth the transitions between regions that have differing volumes as well, but, remember that if you performed a Normalize or Gain function, a new soundfile was written,, and there is no sound beyond the boundary of the region that was adjusted. Therefore, there's nothing to crossfade with beyond the edit point. You can fix this by always using the trimmer tool to extend the edges of a region before performing a gain function. After changing the gain, you can trim the boundaries back to the original locations, but there are now small hidden "handles" which will allow you to crossfade with adjacent regions, or to nudge or trim the new region with greater flexibility. It's often helpful to extend the length of a region when performing any function that writes a new file, allowing greater flexibility in later editing.

Heads and Tails

The very beginnings and endings of voice tracks are often problematic as well. There may be a click at the region boundary, out-of or into silence, or there may be ambience behind the voice that makes a hard edit sound clunky. Sometimes even subtle, low-level background sounds can make a hard edit just sound "wrong" even if there's not an obvious ambience shift. Never fear, fades can help you here too.

Use the Trimmer to adjust the end of the region beyond the last sound, with as much ambience as possible (you might edit some similar ambience onto the end of your region if you don't have enough in the clear in the original soundfile.) The length of the ambience will vary from case to case. For simple smoothing of very quiet ambience, as little as a quarter of a second will suffice. Louder ambience may require a longer section to make a smooth fade. With the Region tool, click once at the end of the voice (you can even set this point just barely in the sound - experiment a bit.)

Choose "Edit>>Fades>>Fade to End" (keyboard shortcut Option G) and the default fade-out will be applied from the point you selected to the end of the region.

(You can adjust your default fade-in, fade-out, and crossfade under "Setups>>Preferences>>Editing" - I recommend an "S Curve" for ins and outs)

This same technique can be used to create a fade-in. The keyboard shortcut is: Option D.

You should leave the creation of crossfades and fade-ins and outs until after you have normalized, or performed any audiosuite gain functions. Those earlier decisions will have a large effect on what fades, if any, you will need to use. If you need to, you can apply gain adjustments to sections with fades, but be sure to select both the audio region AND the fade region, and process them identically (choose "entire selection" rather than "region by region" in your normalize or gain window.) Do not process them individually. The fade is in independent region, and likely has a different maximum peak level from the region to which it is attached, and normalizing it on its own will often make the fade sound wrong.

You can, of course, still make level adjustments for a mix through volume automation or the riding faders in the mixer.

Linguistics for Editing

All those old tricks that many of us remember from reel-to-reel tape days still work in digital editing, and are actually easier to implement. It's easiest to edit on "unvoiced" sounds: sounds that are produced without the vocal chords vibrating. Stops like "t""k" or "p," slushy or fricative sounds like "shhh" "fff" or "th" (in "think", but not "these") give you tremendous flexibility. Sometimes editing to the middle of these sounds, rather than at word boundaries, gives a better result.

In this example, the speaker did a little stutter-repeat, but also slid all the words together, making it hard to cut at the edge of a word. So instead, we'll select from the middles of the "sss" in "so".

Then in shuffle, delete the repeated phrase.

In these sounds, especially "sss" and "ffff," it's hard to make a bad edit.

In voiced sounds, it's trickier, because there is a more regular waveform pattern, so one needs to be more careful in making such edits.

But it's still possible, just try to cut at the same point in the word, and at the same point in the wave-cycle, ideally at a zero-crossing, but it's not crucial. Here the edit was not made at a zero, rather at a peak, but it is a spot that would make a smooth transition.

In voiced sounds, it's trickier, because there is a more regular waveform pattern, so one needs to be more careful in making such edits.

Zero-Crossings are safest for avoiding clicks or thumps, but often editing right from the peak of a "k" or "t" or "p" to another of the same sound is easier, and just as smooth. Or try a small fade across only a few cycles of the waveform.

I've heard that many people are still cutting dialog in other programs and then importing the tracks into ProTools for mixing. There's nothing wrong with that, but these techniques should get you moving quickly when cutting dialog in ProTools free, and may help make a transition to doing all your work within this one program.

~Jeff Towne
Tools Editor, Transom

Coming Soon:
Asymmetrical fades and other fun tricks in the fade window.