Tutorial: Sound Removal, Auto-Speech Alignment, and More in Adobe Audition CC

Working With Automatic Speech Alignment

If you work with scripted projects that need to be delivered with no mistakes whatsoever in the dialogue, ADR, or Additional Dialogue Recording (or Replacement), is something that you're probably familiar with. Adobe has a tool that they introduced in the last version of the Creative Suite that carries over into Audition Creative Cloud called Automatic Speech Alignment.

Let’s say you have a set of clips, and one of them was badly recorded or has some unwanted noise, or one includes dialogue from an actor who appears in multiple clips, but his voice sounds substantially different in one of them To clean up the project, you need to match up the bad recording to a studio version, recorded later, that you want lip-synced to it. Auto Speech Alignment can help.

In the example shown in Figure 16 (below), created from files provided by Adobe for this tutorial, I've got two tracks. On top is the original audio with the video, and you can tell it's varied wildly. There are three different scenes--one outside of the Adobe offices in Seattle, one in a construction zone, and a third that looks like it's outdoors, perhaps near a fountain or a public area, with a lot of traffic and other noises in the background. And you can just look at the waveform in Figure 16 and see the differences from scene to scene. The one outside of the office building is pretty clean. There's a little bit of noise at the end. The one in the construction area has a lot of ambient sound. And then the last one has kind of an in-between range, particularly in the first half of the clip. There's some noise, but it's not quite as bad as the middle one.

Figure 16. Our original audio (top track) and studio re-record (below).

On the contrary, on the track below, we have their studio version, where they brought Durin (the speaker) in to re-record his dialogue. And he recorded a clean version in their nice quite studio so that they'd have a nice clean version of what he needed to say. Listen to the original, just so you get an idea of what we're dealing with, at the 9:00 mark of the clip below.

So there's a lot to deal with here. Fortunately, with Automatic Speech Alignment, all you have to do is select one and two that match. And you can see they recorded just the audio needed for each one of these clips individually. So all you've got to do is get these two clips selected and you've got the same words in each. Then right-click on either one of them, and go down to Automatic Speech Alignment (Figure 17, below).

Figure 17. Choosing Automatic Speech Alignment.

The Automatic Speech Alignment dialog opens (Figure 18, below). There, you're going to choose the reference clip, which is the clip it's going to look at to find out where the words and syllables should go. In this case, it's going to be the top one, the MVI file. And you're also going to choose a channel. Stereo could be used in a variety of ways. But in this case, it looks like they had an on-camera stereo mic. So it really doesn't matter which channel we choose. The channels in the original look about the same. You'd normally want to try to choose the cleaner one so it has an easier time of reading it. But we'll just leave it off. Their studio recording, of course, is in mono. So there's no selection for the channel there. And there are three different settings that you can choose for how it aligns. Each one of them may have different results, depending on what quality your audio is. For now, I'm just going to choose kind of the middle of the road safe one, Balanced Alignment and Stretching, and leave it at that.

Figure 18. Choosing settings in the Automatic Speech Alignment dialog.

When you’re done choosing settings, click OK. Audition is going to analyze it and create a new track with just that synchronized audio. And we're going to solo that. So we only hear the new audio. At the 10:46 mark in the clip below, you can hear how close we got it.

As you can hear, it started off with good intentions and kind of crashed and burned at the end. But let's give it another chance. Let's choose the next two clips (the construction site clip and the corresponding studio clip). Right-click, choose Automatic Speech Alignment. Then choose our source. Again, the channel doesn't really matter because they're both pretty awful. And this one's so bad I'm going to choose reference clip as noisy (Figure 19, below), which I assume will help it filter that out all that noise a little bit better.

Figure 19. Choosing settings for matching the second clip.

Durin is really hard to hear in this clip. So let's just hear what it does. Again, it's going to align it and create a new track. In Figure 20 (below), you can already see that it's improved because the waveforms are now in about the same place. And it appears to be a little off at the end, just visually looking at it.

Figure 20. The second clip, well aligned now.

But let's hear what it sounds like by listening to the clip below at the 11:53 mark in the clip below.

That one was just about spot on. We’ll give it one more shot on this last pair. You know the drill: Right-click. Auto Speech Alignment. It's already choosing the right track. Again, left and right are about the same so we'll leave that. Let's leave this reference clip as noisy, because I heard a diesel idling in the background. Solo the new track, back it up and hear what it sounds like at the 12:41 mark in the clip just above this paragraph.

Wow. I have no idea what happened at the end there. It was right on until the very end. But anyway, you get the idea. Your mileage may vary on this. You have to try various settings. As I showed you, there are three different ways to allow it to automatically align. So one of these might have worked better for one of these pairs of clips. It's all going to vary, depending on your source audio, the quality of it, what kind of noises are in the background, and how consistent it is.

A lot of the renders I did in the video tutorial that accompanies this article were done in real time. Some of them I skipped just for sake of time. But for the most part, they were all done right in front of you on the screen. And I was also in the background recording, in a separate application, my audio and recording the screen capture in Camtasia, which was generously provided by TechSmith. So all of that was eating up some of the RAM, which is why I skipped ahead a few times and cut the video.

But you'll notice right away that the processing of all the effects in real time is extremely fast and vastly improved over previous versions. So that's an overview of some of the newest and improved features in Adobe Audition Creative Cloud.